CN105740815B - A kind of Human bodys' response method based on depth recurrence stratified condition random field - Google Patents
A kind of Human bodys' response method based on depth recurrence stratified condition random field Download PDFInfo
- Publication number
- CN105740815B CN105740815B CN201610064349.0A CN201610064349A CN105740815B CN 105740815 B CN105740815 B CN 105740815B CN 201610064349 A CN201610064349 A CN 201610064349A CN 105740815 B CN105740815 B CN 105740815B
- Authority
- CN
- China
- Prior art keywords
- video
- indicate
- random field
- behavior
- condition random
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- 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
-
- 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
Abstract
The invention discloses a kind of Human bodys' response methods based on depth recurrence stratified condition random field, first, the human body attitude of behavior act main body and the object information that may be interacted with it in the RGB-D video by RGB-D video camera shooting behavior act scene are extracted respectively, using both information as the intermediate layer state of depth recurrence stratified condition random field, the correlation of current state and current all prediction output state set occurred, constructs depth recurrence stratified condition random field models in modeling and forecasting output target-like state layer；Secondly, learning the identification and classification model about human body behavior sequence using the structuring support vector machine classifier of BCFW optimization method driving；Finally, the model parameter obtained according to study and the classification for predicting human body behavior sequence to be tested up to discrimination model.The present invention has significant robustness to behavior act, improves the recognition accuracy of human body behavior act to a certain extent.
Description
Technical field
The present invention relates to a kind of Human bodys' response methods, more particularly to one kind to be based on depth recurrence stratified condition random field
The human body row of (Deep Recursive and Hierarchical Conditional Random Fields, DR-HCRFs)
For recognition methods, belong to computer vision Activity recognition technical field.
Background technique
Human bodys' response has very important position in computer vision, in intelligent monitoring, human-computer interaction and body
It educates and has a wide range of applications in the fields such as video processing.
In recent years, for indoor scene Activity recognition research mainly using probability graph model method to personage's behavior into
Row classification parsing.Common probability graph model is broadly divided into two kinds of structures: generating model and discrimination model.Common generation model
Have: hidden Markov model (Hidden Markov Model), Bayesian network (DBNs), semi-Markov model (Semi-
Markov Models).Generating model needs distribution and correlation to prior information to model, when between the variable of input
There are when more complicated correlation, the modeling of Joint Distribution will become complicated or even inaccuracy.Discrimination model pair is used on the contrary
Conditional probability modeling can derive accurate and effectively infer.Such as: hidden conditional random fields (Hidden CRFs), cyclic annular item
Part random field (Loopy CRFs).
In addition, existing Activity recognition research lays particular emphasis on the addition semantic context information in original technology, such as object
With the contextual information of behavior, behavior and behavior.Being experimentally confirmed these semantic informations can be as hidden in discrimination model
The accuracy of state raising Activity recognition.What O.Sener and A.Saxena.rCRF was delivered on RSS in 2015
" recursive belief estimation over CRFs in RGB-D activity videos ", from rgb video
It extracts human body attitude feature and object enlightens information, the conditional random field models by the way that recursive Bayesian estimation is added calculate people
Relation of trust between body behavior.What Y.Jiang, H.S.Koppula and A.Saxena were delivered on PAMI in 2015
" Modeling 3D environments through hidden human context " proposes a kind of infinite hidden state
Conditional random field models, to human postures a large amount of in 3D environment and interaction object processing have obvious action.
" the The infinite-order conditional that S.P.Chatzis and Y.Demiris was delivered on PAMI in 2013
Random field model for sequential data modeling ", propose a kind of Infinite Order condition random field pair
Serialized data modeling, while having used in the Method Modeling sequence label of serial memorization (sequence memorizer) a kind of
The correlation of Infinite Order.
The above-mentioned existing Activity recognition method based on probability graph model does not all consider the inside table of dbjective state simultaneously
Show the higher order dependencies between state, however it remains the low problem of recognition accuracy.
Summary of the invention
The technical problems to be solved by the present invention are: providing a kind of human body row based on depth recurrence stratified condition random field
For recognition methods, using human body attitude and interaction object as the intermediate representation state of prediction dbjective state, building one comprising defeated
Enter the depth recurrence stratified condition random field models of data, intermediate state and target prediction state.
The present invention uses following technical scheme to solve above-mentioned technical problem:
A kind of Human bodys' response method based on depth recurrence stratified condition random field, includes the following steps:
Step 1, the RGB-D training video sample of human body behavior is obtained, which includes rgb video
Information, depth information and human skeleton information combine rgb video information and human skeleton information, and therefrom extract human body attitude
The relative seat feature of feature, the shape of interaction object and position feature and human body and interaction object, features described above is connected
After obtain behavior representation feature；
Step 2, the behavior representation feature obtained according to step 1 constructs behavior representation feature, human body appearance in current video section
Full-mesh probability graph model made of state and the intermediate state of interaction object composition, the link of behavior prediction label three parts, in conjunction with
In training video sample first video-frequency band to current video section previous video section behavior prediction label, foundation work as forward sight
The depth recurrence stratified condition random field models of frequency range；
Step 3, using mean field approximation algorithm, the depth recurrence stratified condition random field models that step 2 is established are converted
For first-order linear chain condition random field models；
Step 4, using maximum-interval arithmetic, the ginseng for the first-order linear chain condition random field models that learning procedure 3 obtains
Number；
Step 5, the first-order linear chain condition random field models and step 4 obtained according to step 3 learn obtained parameter, know
The corresponding behavior prediction label of other test video sample.
Preferably, potential-energy function Ψ (y, h, o, the x of the depth recurrence stratified condition random field models；ω) are as follows:
Wherein, t=1 ..., T indicate t-th of video-frequency band of training video sample, ω1、ω2、ω3、ω4Indicate model
Parameter, ht、ot、ytRespectively indicate human body attitude, the interaction object, behavior prediction label of t-th of video-frequency band；Indicate xtAnd ht、otDependence, φ (xt) indicate t-th of video-frequency band in behavior representation feature xtTo spy
Levy the mapping function in space；Indicate htAnd otBetween correlation,Indicate interaction object s the
Whether appeared in action process in t video-frequency band,Indicate the set of all interactive objects in t video-frequency band, S indicates training view
The set of all interactive objects in frequency sample；ω3(yt,ht,ot) indicate ytAnd ht、otCoupling；It indicates
Historical setWith ytCorrelation.
Preferably, the detailed process of the step 3 are as follows: find out the optimum behavior prediction label of current video section, it is described most
Excellent behavior prediction labelIt may be expressed as:
Wherein,Table
Show the 1st video-frequency band of training video sample to the optimum behavior prediction label of the t-2 video-frequency band, v, u indicate candidate row
For the candidate behavior prediction label in prediction label set Y={ 1 ... V }, V indicates the candidate behavior prediction mark of composition set Y
The total number of label.
Preferably, the calculation expression of the parameter are as follows:Wherein, λ indicates balanced power
Weight values, ω indicate the parameter of model,Indicate the optimum behavior prediction label of i-th of training video sample, N indicates training video
The total number of sample,Indicate the optimum behavior prediction label of i-th of training video sampleWith agenda label
yiThe loss function of difference.
Preferably, the tool that the RGB-D training video sample of human body behavior is obtained described in step 1 is Kinect depth sensing
Device.
The invention adopts the above technical scheme compared with prior art, has following technical effect that
The present invention is based on the Human bodys' response methods of depth recurrence stratified condition random field, introduce and increase in modeling
The higher order dependencies between intermediate structure and the past behavior act inside current behavior movement correlative factor, to behavior act
Body configuration's difference, complex scene and interaction object for including in the process etc. all have significant robustness, to a certain extent
The recognition accuracy of human body behavior act can be improved.
Detailed description of the invention
Fig. 1 is the schematic diagram of depth recurrence stratified condition random field proposed by the present invention.
Fig. 2 is that the present invention is based on the flow charts of the Human bodys' response method of depth recurrence stratified condition random field.
Specific embodiment
Embodiments of the present invention are described below in detail, the example of the embodiment is shown in the accompanying drawings.Below by
The embodiment being described with reference to the drawings is exemplary, and for explaining only the invention, and is not construed as limiting the claims.
The present invention is based on detailed process such as Fig. 2 institutes of the Human bodys' response method of depth recurrence stratified condition random field
Show, specifically according to the following steps:
Step 1, the RGB-D video sequence that human body behavior is captured using Kinect depth transducer, by acquiring
The depth information of scene is taken the photograph, to extract the human skeleton structural information of behavior act Subject-Human, and combines human skeleton and RGB
Two kinds of data sources of video sequence extract wherein human body attitude feature, the shape of interaction object and position feature and human body and object
The relative position information of body, and these information expression of connecting, combination form the behavior representation feature finally observed, and as rear
Continuous input.
Step 2, basis are up to the behavior representation feature of input, and building is by input observation data xtTo human body attitude htAnd friendship
Mutual object otThe intermediate state of composition and final behavior prediction label ytFull-mesh probability graph model made of three parts link,
Establish corresponding depth recurrence stratified condition random field models.
Data x is observed for input, corresponding behavior label is that the probability of y can be expressed as objective function:
Wherein, Ψ (y, h, o, x；ω) the potential-energy function of depth recurrence stratified condition random field models measures input observation
Data x, the correlation in middle layer between human body attitude h and interaction object o and behavior prediction label y, ω are that model is joined
Number.Potential-energy function Ψ (y, h, o, x；Include ω) four component parts, is respectively as follows:
A. input observation data xtWith intermediate state htAnd otDependence:
Ψ1(ht, ot, xt；ω1)=ω1(ht, ot)·φ(xt) (2)
Wherein, φ (xt) it is the mapping function for inputting data into feature space.
B. human body attitude htWith interaction object otBetween correlation:
Wherein,Indicate whether object s appears in action process in t video-frequency band, whenWhen table
It shows existing；WhenExpression does not occur.The set of S expression all objects label.
C. behavior prediction label ytWith intermediate state htAnd otCoupling:
Ψ3(yt,ht,ot；ω3)=ω3(yt,ht,ot) (4)
D. the historical set of dbjective stateWith preceding dbjective state behavior prediction label ytCorrelation:
For Current observation to input observation data x (assuming that T indicate the input data sequence total length or training view
The total video section of frequency sample), in conjunction with above-mentioned formula (2)-(5), the gesture of available depth recurrence stratified condition random field models
Energy function, is specifically expressed as follows:
Step 3, using mean field approximation algorithm (mean-field-like), to random up to depth recurrence stratified condition
Field carries out model inference, to reduce the computation complexity about constructed model, is converted to first-order linear chain condition random field mould
Type.
The purpose of model inference is in the case where given graph model and parameter ω, and simulated target letter can be maximized by finding out
Several y:
Formula (6) is changed into the dynamic programming problems of backward recursive:
Wherein, v indicates prediction dbjective state, is an element in target collection Y={ 1 ... V }, and V indicates composition set
The element total number of Y.
Original state are as follows:
ζ1(v)=ω3(y1=v, h1, o1) (10)
By ζ1(v) definition is it is found that the dynamic programming problems can be regarded as the feelings in known historical forecast optimum state
Under condition, the corresponding behavior label of prediction present input data.Using original state as historic state, in available video sequence
The corresponding ζ of second time periodt(v):
According to the first and second periods, third period corresponding ζ can be further obtainedt(v):
Herein, present invention assumes that ytState value not by recursive expression ζ3(v) y in2And y3It influences, works as y1It obtains optimal
Predicted state valueWhen, above formula can further indicate that are as follows:
As ζ3(v) mean field approximation result.Similarly in the ζ of known 4th periodt(v) are as follows:
By the angle of mean field approximation algorithm, ignore y1And y2Influence of fluctuations:
By the near-optimal method, the dynamic programming problems in the present invention can be changed into one with depth recursive function
ζt(v) simple problem with worst upper boundary of calculation complexity indicated.ζt(v) the Infinite Order correlation for including in may be expressed as:
Step 4, using maximum-interval arithmetic (Max-margin), learn the parameter of constructed random field models.
In training datasetIn include N group training video sample xiWith behavior prediction sequence label yi
One-to-one sample.But intermediate state h and o is unknown, and the purpose of model learning is to find out optimal model parameter in the present invention
ω, so that the difference of predictive behavior note and practical note is minimum.To prevent over-fitting, the present invention provides corresponding canonical
Item expression formula:
Wherein λ is equalizing weight value,It indicates by formula (7) optimum behavior prediction label sequence obtained.To indicate to predict the loss function of output behavior sequence and agenda sequence difference.It is embodied as:
Wherein z is in given sequence length ZiUnder, z-th of value in sequence.Directly solving to formula (18) is a N-P difficulty
Topic, the present invention carry out marginalisation substitution to primary loss function, find out the coboundary of given loss function.So by primal problem
It is rewritten as with the minimum problems for solving objective function under constraint:
Wherein differenceSlack variable ξiIt indicates i-th
The substitution loss function of data point.A usable structuring support vector machines is converted by primal problem herein
The convex problem that (Structural-Support Vector Machine, SSVM) is solved.But due to YiInternal various associativities
Matter, leads to the convex problem there are exponential constraint, and in learning process the time spent in is larger.The present invention uses N number of
Piecewise linearity is to substitution Σi|Yi| linear restriction, the corresponding structuring chain type loss function of defined formula (19):
Finally obtain the equivalent no constraint formulations of formula (17):
Finally by original-antithesis Frank-Wolf (the block-coordinate primal-dual of block-coordinate
Frank-Wolfe, BCFW) algorithm solves the result of the problem.
Step 5, the model parameter obtained according to depth recurrence stratified condition random field models and study, to test data set
Predict behavior label corresponding to each video sequence.
The above examples only illustrate the technical idea of the present invention, and this does not limit the scope of protection of the present invention, all
According to the technical idea provided by the invention, any changes made on the basis of the technical scheme each falls within the scope of the present invention
Within.
Claims (5)
1. a kind of Human bodys' response method based on depth recurrence stratified condition random field, which is characterized in that including walking as follows
It is rapid:
Step 1, obtain human body behavior RGB-D training video sample, the RGB-D training video sample include rgb video information,
Depth information and human skeleton information combine rgb video information and human skeleton information, and therefrom extract human body attitude feature,
The shape and position feature and human body of interaction object and the relative seat feature of interaction object, after features described above is connected
To behavior representation feature；
Step 2, the behavior representation feature obtained according to step 1, construct current video section in behavior representation feature, human body attitude and
Full-mesh probability graph model made of intermediate state, the behavior prediction label three parts of interaction object composition link, combined training
In video sample first video-frequency band to current video section previous video section behavior prediction label, establish current video section
Depth recurrence stratified condition random field models；
Step 3, using mean field approximation algorithm, the depth recurrence stratified condition random field models that step 2 is established are converted to one
Rank linear chain conditional random field model；
Step 4, using maximum-interval arithmetic, the parameter for the first-order linear chain condition random field models that learning procedure 3 obtains；
Step 5, the first-order linear chain condition random field models and step 4 obtained according to step 3 learn obtained parameter, and identification is surveyed
Try the corresponding behavior prediction label of video sample.
2. the Human bodys' response method as described in claim 1 based on depth recurrence stratified condition random field, which is characterized in that
Potential-energy function Ψ (y, h, o, the x of the depth recurrence stratified condition random field models；ω) are as follows:
Wherein, t=1 ..., T indicate t-th of video-frequency band of training video sample, ω1、ω2、ω3、ω4Indicate the ginseng of model
Number, ht、ot、ytRespectively indicate human body attitude, the interaction object, behavior prediction label of t-th of video-frequency band；Table
Show xtAnd ht、otDependence, φ (xt) indicate t-th of video-frequency band in behavior representation feature xtTo the mapping letter of feature space
Number；Indicate htAnd otBetween correlation,Indicate whether interaction object s goes out in t video-frequency band
In present action process,Indicate the set of all interactive objects in t video-frequency band, S indicates all friendships in training video sample
The set of mutual object；ω3(yt,ht,ot) indicate ytAnd ht、otCoupling；Indicate historical setWith
ytCorrelation.
3. the Human bodys' response method as described in claim 1 based on depth recurrence stratified condition random field, which is characterized in that
The detailed process of the step 3 are as follows: find out the optimum behavior prediction label of current video section, the optimum behavior prediction label
It may be expressed as:
Wherein, Indicate instruction
Practice the 1st video-frequency band of video sample to the optimum behavior prediction label of the t-2 video-frequency band, v, u indicate that candidate behavior is pre-
The candidate behavior prediction label in tag set Y={ 1 ... V } is surveyed, V indicates the candidate behavior prediction label of composition set Y
Total number, ω1、ω2、ω3、ω4Indicate the parameter of model, ht、ot、ytIt respectively indicates the human body attitude of t-th of video-frequency band, hand over
Mutual object, behavior prediction label,Indicate xtAnd ht、otDependence,Indicate htAnd ot
Between correlation,Indicate whether interaction object s appears in action process in t video-frequency band,Indicate t
The set of all interactive objects in video-frequency band, S indicate the set of all interactive objects in training video sample；ω3(yt=v, ht,
ot) indicate ytAnd ht、otCoupling；It indicatesWith yt、yt-1Correlation.
4. the Human bodys' response method as described in claim 1 based on depth recurrence stratified condition random field, which is characterized in that
The calculation expression of the parameter are as follows:
Wherein, λ indicates equalizing weight value, and ω indicates the parameter of model,Indicate that the optimum behavior of i-th of training video sample is pre-
Mark label, N indicate the total number of training video sample,Indicate the optimum behavior prediction of i-th of training video sample
LabelWith agenda label yiThe loss function of difference.
5. the Human bodys' response method as described in claim 1 based on depth recurrence stratified condition random field, which is characterized in that
The tool that the RGB-D training video sample of human body behavior is obtained described in step 1 is Kinect depth transducer.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610064349.0A CN105740815B (en) | 2016-01-29 | 2016-01-29 | A kind of Human bodys' response method based on depth recurrence stratified condition random field |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610064349.0A CN105740815B (en) | 2016-01-29 | 2016-01-29 | A kind of Human bodys' response method based on depth recurrence stratified condition random field |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105740815A CN105740815A (en) | 2016-07-06 |
| CN105740815B true CN105740815B (en) | 2018-12-18 |
Family
ID=56247141
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610064349.0A Active CN105740815B (en) | 2016-01-29 | 2016-01-29 | A kind of Human bodys' response method based on depth recurrence stratified condition random field |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105740815B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107341471B (en) * | 2017-07-04 | 2019-10-01 | 南京邮电大学 | A kind of Human bodys' response method based on Bilayer condition random field |
| CN107491735B (en) * | 2017-07-20 | 2020-08-18 | 浙江工业大学 | Tag and interaction relation joint learning method for human behavior recognition |
| CN111914807B (en) * | 2020-08-18 | 2022-06-28 | 太原理工大学 | Miner behavior identification method based on sensor and skeleton information |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103810496A (en) * | 2014-01-09 | 2014-05-21 | 江南大学 | 3D (three-dimensional) Gaussian space human behavior identifying method based on image depth information |
| CN104598890A (en) * | 2015-01-30 | 2015-05-06 | 南京邮电大学 | Human body behavior recognizing method based on RGB-D video |
| CN104809187A (en) * | 2015-04-20 | 2015-07-29 | 南京邮电大学 | Indoor scene semantic annotation method based on RGB-D data |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7526123B2 (en) * | 2004-02-12 | 2009-04-28 | Nec Laboratories America, Inc. | Estimating facial pose from a sparse representation |
-
2016
- 2016-01-29 CN CN201610064349.0A patent/CN105740815B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103810496A (en) * | 2014-01-09 | 2014-05-21 | 江南大学 | 3D (three-dimensional) Gaussian space human behavior identifying method based on image depth information |
| CN104598890A (en) * | 2015-01-30 | 2015-05-06 | 南京邮电大学 | Human body behavior recognizing method based on RGB-D video |
| CN104809187A (en) * | 2015-04-20 | 2015-07-29 | 南京邮电大学 | Indoor scene semantic annotation method based on RGB-D data |
Non-Patent Citations (1)
| Title |
|---|
| 人体行为识别的条件随机场方法;王媛媛 等;《重庆理工大学学报(自然科学)》;20130630;第27卷(第6期);第93-99、105页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105740815A (en) | 2016-07-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Song et al. | Pedestrian trajectory prediction based on deep convolutional LSTM network | |
| Kong et al. | Integrated generative model for industrial anomaly detection via bidirectional LSTM and attention mechanism | |
| CN112766244B (en) | Target object detection method and device, computer equipment and storage medium | |
| CN103003846B (en) | Articulation region display device, joint area detecting device, joint area degree of membership calculation element, pass nodular region affiliation degree calculation element and joint area display packing | |
| CN107886069A (en) | A kind of multiple target human body 2D gesture real-time detection systems and detection method | |
| Cheng et al. | Bi-pointflownet: Bidirectional learning for point cloud based scene flow estimation | |
| CN105320944A (en) | Human body behavior prediction method based on human body skeleton movement information | |
| CN110135249A (en) | Human bodys' response method based on time attention mechanism and LSTM | |
| CN103544496A (en) | Method for recognizing robot scenes on basis of space and time information fusion | |
| CN111028319B (en) | Three-dimensional non-photorealistic expression generation method based on facial motion unit | |
| Yu et al. | Human motion based intent recognition using a deep dynamic neural model | |
| Obo et al. | Localization of human in informationally structured space based on sensor networks | |
| CN105740815B (en) | A kind of Human bodys' response method based on depth recurrence stratified condition random field | |
| Lu et al. | MGRL: Graph neural network based inference in a Markov network with reinforcement learning for visual navigation | |
| Fang et al. | Dynamic gesture recognition using inertial sensors-based data gloves | |
| CN107341471B (en) | A kind of Human bodys' response method based on Bilayer condition random field | |
| CN113821724B (en) | Time interval enhancement-based graph neural network recommendation method | |
| Mei et al. | Differential reinforcement and global collaboration network for rgbt tracking | |
| Wu et al. | Vision-language navigation: a survey and taxonomy | |
| Zhou et al. | Time-varying trajectory modeling via dynamic governing network for remaining useful life prediction | |
| Liu et al. | Human motion target recognition using convolutional neural network and global constraint block matching | |
| CN114168769B (en) | Visual question-answering method based on GAT relation reasoning | |
| Bai et al. | Real-time 3d human pose estimation without skeletal a priori structures | |
| CN115204171A (en) | Document-level event extraction method and system based on hypergraph neural network | |
| Kim et al. | Dynamic arm gesture recognition using spherical angle features and hidden markov models |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| EE01 | Entry into force of recordation of patent licensing contract | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20160706 Assignee: Nanjing Nanyou Information Industry Technology Research Institute Co. Ltd. Assignor: Nanjing Post & Telecommunication Univ. Contract record no.: X2019980001257 Denomination of invention: Human body behavior identification method based on deep recursive and hierarchical condition random fields Granted publication date: 20181218 License type: Common License Record date: 20191224 |
|
| EE01 | Entry into force of recordation of patent licensing contract | ||
| EE01 | Entry into force of recordation of patent licensing contract |
Application publication date: 20160706 Assignee: Jiangsu Tuoyou Information Intelligent Technology Research Institute Co.,Ltd. Assignor: NANJING University OF POSTS AND TELECOMMUNICATIONS Contract record no.: X2021320000043 Denomination of invention: A human behavior recognition method based on deep recursive hierarchical conditional random field Granted publication date: 20181218 License type: Common License Record date: 20210616 |
|
| EC01 | Cancellation of recordation of patent licensing contract | ||
| EC01 | Cancellation of recordation of patent licensing contract |
Assignee: NANJING NANYOU INSTITUTE OF INFORMATION TECHNOVATION Co.,Ltd. Assignor: NANJING University OF POSTS AND TELECOMMUNICATIONS Contract record no.: X2019980001257 Date of cancellation: 20220304 |