CN103150752A - Key mark point-based human body posture sparse reconstruction method - Google Patents

Key mark point-based human body posture sparse reconstruction method Download PDF

Info

Publication number
CN103150752A
CN103150752A CN 201310047891 CN201310047891A CN103150752A CN 103150752 A CN103150752 A CN 103150752A CN 201310047891 CN201310047891 CN 201310047891 CN 201310047891 A CN201310047891 A CN 201310047891A CN 103150752 A CN103150752 A CN 103150752A
Authority
CN
China
Prior art keywords
attitude
database
control signal
data
articulation point
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
Application number
CN 201310047891
Other languages
Chinese (zh)
Other versions
CN103150752B (en
Inventor
肖俊
林海
庄越挺
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Zhejiang University ZJU
Original Assignee
Zhejiang University ZJU
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Zhejiang University ZJU filed Critical Zhejiang University ZJU
Priority to CN201310047891.1A priority Critical patent/CN103150752B/en
Publication of CN103150752A publication Critical patent/CN103150752A/en
Application granted granted Critical
Publication of CN103150752B publication Critical patent/CN103150752B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Landscapes

  • Processing Or Creating Images (AREA)

Abstract

The invention discloses a key mark point-based human body posture sparse reconstruction method. The method comprises the following steps of: extracting articulation point information corresponding to a key mark point to construct a k-d tree index for human body exercise data in a database; looking up k neighbors which are most matched with a control signal input by each frame in the database for the control signal input by each frame; constructing a recovery dictionary by the neighbors; and recovering the exercise posture of the whole body by utilizing a sparely expressed optimal frame and incorporating a result recovered by a previous frame into smoothness constraint of the current frame, wherein the recovered missing key point information is natural enough and no jitter exists in time sequence.

Description

A kind of human body attitude sparse reconstruction method based on crucial gauge point
Technical field
The present invention relates to three-dimensional graphics and virtual reality field, relate in particular to a kind of human body attitude sparse reconstruction method.
Background technology
The human motion capture technology has been widely applied to cartoon making, the film special efficacy, and the entertainment field such as computer game are brought huge economic benefit.Although human motion be followed the tracks of and be recorded to present various business capture systems can with very high spatial and temporal resolution; but because the very expensive software of needs, equipment and data acquisition (are dressed pretty troublesome; business machine usually need the user dress 40 ?50 reflective marker points that carefully are placed with; perhaps 18 magnetic force or inertial sensor); in addition; they can have special requirement and restriction to performer and the collection environment of performance usually, cause them to be not suitable for family expenses.
In using, some home entertainings do not need to capture very accurate human body attitude, often only need to know the mutual user of current performance be walk, be to run, be to squat or be to jump, only need to know hand and pin be arranged in the space general area and and be indifferent to the accurate coordinates of finger, toe, ancon, knee, at this moment, use accurate commercial capture device just obviously to break a fly upon a wheel.Demand to the motion capture system of non-precision cheaply is day by day strong, the sensor that makes it possible to by seldom becomes very attractive with regard to the system that extracts the significant movable information of user, takes off time of wearing and the complexity of data acquisition because they have greatly reduced equipment.Although the concrete technological means that these systems realize is had nothing in common with each other, the thought on main body is exactly to utilize the low-dimensional control signal of catching to recover human body attitude.Therefore, the high-dimensional athletic posture of natural reality has challenging problem with regard to being called one how to utilize the control signal of hanging down dimension to recover as far as possible.
When utilizing the control signal of low-dimensional to catch all-around exercises, have many dimensions be do not have bound, the data of these dimensions or need to be calculated according to the method for inverse kinematics, or need to be synthesized or rebuild out according to data in database.Traditional inverse kinematics (Inverse Kinematics), in the situation that known human body endpoint node position, the nonlinear system of utilizing the kinematics Formula Solution to owe fixed calculates the parameter configuration of human body institute related node, even if added the constraint of anthropometry aspect, the computation process of finding the solution of inverse kinematics remains very complicated, and the attitude of finally calculating may be natural not.At this moment, the method take the exercise data of catching in advance as the data-driven that supports has just shown power.
In recent years, the denoising method that based on data drives is risen gradually, the pure exercise data that these method utilizations have been caught is as the database support, with the control signal of current input as querying condition, search neighbour's attitude of mating the most with it, utilize these neighbour's attitudes to be reconstructed, the difference of each method is mainly reflected in how above reconstruct.This method makes exercise data to reuse, and has further reduced industrial cost and the cost performance that has improved capturing movement.
Summary of the invention
The present invention is achieved through the following technical solutions: a kind of 3 d human motion attitude sparse reconstruction method based on crucial gauge point is provided, comprises the steps:
(1) the pure human body movement data of download storehouse;
(2) use capturing movement equipment to catch human body movement data: we wish to recover with the least possible articulation point information the attitude of whole body, intuition from the people, four limbs and head all need to keep an endpoint node at least, in addition we need to follow the tracks of the root node that is affixed on buttocks global position and towards, with convenient, input signal is carried out translation and rotation.Therefore, use the Vicon motion capture system that the trace information of the reflective marker point that is affixed on this main points joint of human body is caught;
(3) data that step 1 obtained are carried out pre-service: be the TRC form with the data of database in step 1 from the BVH format conversion, simultaneously each attitude carried out translation, rotational transform; Translation transformation is mainly that the root node motion that is positioned at buttocks of the exercise data initial point to global coordinate system is gone, and rotational transform is that the normal vector on the plane (by marker point institute's match plane out of pasting on health) at body trunk place is alignd with x coordinate in global coordinate system, guarantee all attitudes after processing have identical position and towards; Write down transform matrix M Trans, its contravariant is changed to
Figure BDA00002811784800021
(4) translation and the postrotational exercise data index building step 3 exported: for the control signal that makes input can arrive with it the athletic posture of coupling the most by fast finding in database, we are to the data after processing through step 3, extract these attitudes articulation point information corresponding with input control signal, be that head, left finesse, right finesse, left heel and right crus of diaphragm are followed 5 articulation points, the data construct k-d of totally 15 dimensions sets index.
(5) search the athletic posture that mates most with input control signal in database: the control signal of constantly inputting for t
Figure BDA00002811784800022
Wherein k is for controlling the number of articulation point, and 3 in 3*k represents x, y, three coordinate dimensions of z; We utilize the k-d tree to search n the proper vector that approaches the most with it in database Define the matrix of this n neighbour's feature composition for recovering dictionary
Figure BDA00002811784800024
Database attitude submatrix corresponding to these features is
Figure BDA00002811784800025
(6) these a small amount of articulation point information are carried out linear regression based on neighbour's attitude: the attitude that we wish to recover is by this P tIn linear composition of attitude, find the solution reconstructed coefficients ω so problem is converted into t, make the attitude of recovery
Figure BDA00002811784800031
The coefficient solution formula is the energy minimization equation, is defined as:
ω t ^ = arg min ω t ( E control ( ω t ) + λ 2 E smooth ( ω t ) ) (formula 1)
Wherein
Figure BDA00002811784800033
Be the reconstructed coefficients that we will find the solution, E Control(ω t) the expression control item, i.e. the attitude that assurance recovers is wanted the constraint requirements of As soon as possible Promising Policy low-dimensional control signal, E Smooth(ω t) the level and smooth item of expression, the constraint iff considering current control signal may make the frame of recovery and the shake that high frequency appears in interframe, adds constraint by the attitude that former frame is recovered, and can guarantee the flatness requirement, λ 2The weight that expression is level and smooth.Particularly, we define:
E control ( ω t ) | | Ds t * ω t - y t | | 2 2 + λ 1 | | ω t | | 1
Y wherein tBe current time low-dimensional control signal,
Figure BDA00002811784800035
The control that a small amount of articulation point data of serving as reasons consist of recovers dictionary, λ 1Represent sparse coefficient, when a small amount of articulation point corresponding to the attitude that guarantees to recover approaches as far as possible with control signal, the sparse property of assurance reconstructed coefficients.
Figure BDA00002811784800036
Wherein
Figure BDA00002811784800037
Be the whole body attitude that former frame recovers, P tThe pose recovery dictionary that neighbour's attitude that expression is indexed by control signal consists of by the former frame attitude, can retrain non-control articulation point.We user finds the solution (formula 1) to alternative algorithm.Recover the present frame attitude:
Figure BDA00002811784800038
(7) calculate the world coordinates that recovers attitude: according to the root node information of catching in step 2, the attitude that recovers is carried out translation and rotate obtaining its position under global coordinate system, be net result.
The invention has the beneficial effects as follows, the method of neighbour's athletic posture is mated in the present invention the most by based on data library lookup and control signal, computation process and the natural not result of inverse kinematics complexity have not only been avoided, and the exercise data of catching before making has obtained reusing, and has reduced industrial cost.The Optimization Framework based on sparse expression of the present invention by proposing, and the attitude that former frame is recovered as level and smooth according to adding consideration, make the athletic posture that recovers very occur shaking on nature and sequential.
Description of drawings
Fig. 1 is based on the process flow diagram of the 3 d human motion attitude sparse reconstruction method of data-driven.
Embodiment
Describe the present invention in detail below in conjunction with accompanying drawing, it is more obvious that purpose of the present invention and effect will become.
As shown in Figure 1, the 3 d human motion attitude sparse reconstruction method that the present invention is based on crucial gauge point comprises the steps:
Step 1: download pure human body movement data storehouse.
Exercise data in database is the pure human body movement data that does not comprise any noise that uses accurate business equipment collection.Can directly use the 3 d human motion data storehouse that CMU provides (can from Http:// mocap.cs.cmu.eduDownload), comprised the various dissimilar motion that gathers from different people in this database, as run, walk, jump and various sports.
Step 2: use capturing movement equipment to catch low-dimensional reference mark signal.
Use Vicon motion capture system ( Http:// www.vicon.com/), trace information to the reflective marker point that is affixed on human body main points joint catches, for with database in data be consistent, we also use CMU(Carnegie Mellon University, CMU) (reference of point set allocation plan Http:// mocap.cs.cmu.edu/markerPlacementGuide.pdf).These six articulation points: root node, head node, left finesse, right finesse, left heel and right crus of diaphragm are followed, and correspond respectively to ROOT, HEAD, LWRIST, RWRIST, LHEEL, RHEEL articulation point in the point set configuration.
Step 3: the data that obtain in step 1 are carried out pre-service.
Use MotionBuilder software that the data of database in step 1 are the TRC form from the BVH format conversion, simultaneously each attitude is carried out translation, rotational transform.Translation transformation is mainly that the ROOT node that is arranged in buttocks with exercise data (configures according to step 2, the ROOT node is affixed on buttocks) initial point that moves to global coordinate system goes, and rotational transform is that the normal vector on the plane (by marker point institute's match plane out of pasting on health) at body trunk place is alignd with x coordinate in global coordinate system, guarantee all attitudes after processing have identical position and towards.Write down transform matrix M Trans, its contravariant is changed to
Figure BDA00002811784800041
Step 4: to translation and the postrotational exercise data index building of step 3 output
For our exercise data, extract corresponding to HEAD, LWRIST, RWRIST, LHEEL, RHEEL articulation point data as index construct k ?the d tree.Allow
Figure BDA00002811784800042
Attitude set in the expression database, m is the frame number of database, d represents the dimension of human body movement data.Matrix P represents database kinematic matrix: P=[p 1, p 2..., p m].We define Φ and are operating as and extract the capable vector of non-zero corresponding to matrix and be combined into submatrix, the definition mask code matrix
Figure BDA00002811784800043
It is all 1 to HEAD, LWRIST, RWRIST, LHEEL, dimension element that the RHEEL articulation point is relevant, and other row elements are 0.Database feature matrix
Figure BDA00002811784800046
Figure BDA00002811784800044
Wherein k=5 * 3=15 represent control signal and k ?the dimension of d tree, at last to D build k ?d tree index.
Step 5: search the athletic posture that mates most with input control signal in database.
Control signal for the input of t frame
Figure BDA00002811784800045
Wherein k is for controlling the number of articulation point, and 3 in 3*k represents x, y, three coordinate dimensions of z; We utilize the k-d tree to search n the proper vector that approaches the most with it in database
Figure BDA00002811784800051
Define the matrix of this n neighbour's feature composition for recovering dictionary
Figure BDA00002811784800052
Database attitude submatrix corresponding to these features is
Figure BDA00002811784800053
Step 6: these a small amount of articulation point information are carried out linear regression based on neighbour's attitude.
The attitude that we wish to recover is by this P tIn linear composition of attitude, find the solution reconstructed coefficients ω so problem is converted into t, make the attitude of recovery
Figure BDA00002811784800054
The coefficient solution formula is the energy minimization equation, is defined as:
ω t ^ = arg min ω t ( E control ( ω t ) + λ 2 E smooth ( ω t ) ) (formula 1)
Wherein Be the reconstructed coefficients that we will find the solution, E Control(ω t) the expression control item, i.e. the attitude that assurance recovers is wanted the constraint requirements of As soon as possible Promising Policy low-dimensional control signal, E Smooth(ω t) the level and smooth item of expression, the constraint iff considering current control signal may make the frame of recovery and the shake that high frequency appears in interframe, adds constraint by the attitude that former frame is recovered, and can guarantee the flatness requirement, λ 2The weight that expression is level and smooth.Particularly, we define:
E control ( ω t ) | | Ds t * ω t - y t | | 2 2 + λ 1 | | ω t | | 1
Y wherein tBe current time low-dimensional control signal, Ds tThe control that a small amount of articulation point data of serving as reasons consist of recovers dictionary, λ 1Represent sparse coefficient, when a small amount of articulation point corresponding to the attitude that guarantees to recover approaches as far as possible with control signal, the sparse property of assurance reconstructed coefficients.
Figure BDA00002811784800058
Wherein
Figure BDA00002811784800059
Be the whole body attitude that former frame recovers, P tThe pose recovery dictionary that neighbour's attitude that expression is indexed by control signal consists of by the former frame attitude, can retrain non-control articulation point.With these two substitution formula, can obtain complete formula:
Wherein
Figure BDA000028117848000511
Can merge and obtain:
Figure BDA000028117848000512
The user finds the solution top formula to alternative algorithm, wherein sparse regular terms parameter lambda 1Be set to 0.1, a level and smooth parameter lambda 2Be set to 0.2.The reconstruction coefficients of finding the solution out multiply by the reconstruct dictionary, the whole body attitude that can be restored out.
Step 7: according to the root node information of catching in step 2, the attitude that recovers is carried out translation and rotate obtaining its position under global coordinate system, be net result.
The method of neighbour's athletic posture is mated in the present invention the most by based on data library lookup and control signal, computation process and the natural not result of inverse kinematics complexity have not only been avoided, and the exercise data of catching before making has obtained reusing, and has reduced industrial cost.The Optimization Framework based on sparse expression of the present invention by proposing, and the attitude that former frame is recovered as level and smooth according to adding consideration, make the athletic posture that recovers very occur shaking on nature and sequential.

Claims (1)

1. the 3 d human motion attitude sparse reconstruction method based on crucial gauge point, is characterized in that, comprises the steps:
Step 1: download pure human body movement data storehouse;
Step 2: use capturing movement equipment to catch low-dimensional reference mark signal: the trace information to the reflective marker point that is affixed on human body main points joint catches, these six articulation points are: root node, head node, left finesse, right finesse, left heel and right crus of diaphragm are followed, and correspond respectively to ROOT, HEAD, LWRIST, RWRIST, LHEEL and RHEEL articulation point in exercise data;
Step 3: the data in the database that obtains in step 1 are carried out pre-service: be the TRC form with the data of the database of download in step 1 from the BVH format conversion, simultaneously each attitude carried out translation and rotational transform;
Translation transformation is mainly that the ROOT node motion that is positioned at buttocks of the exercise data initial point to global coordinate system is gone, and rotational transform is that the normal vector on the plane at body trunk place is alignd with x coordinate in global coordinate system, guarantee all attitudes after processing have identical position and towards; Write down transform matrix M Trans, its contravariant is changed to M trans - 1 ;
Step 4: to the translation of step 3 output and the exercise data index building after rotational transform: to the data after processing through step 3, extract corresponding to HEAD, LWRIST, RWRIST, LHEEL, RHEEL articulation point data as index construct k ?the d tree; Order
Figure FDA00002811784700012
Attitude set in the expression database, m is the frame number of database, d represents the dimension of human body movement data; Matrix P represents database kinematic matrix: P=[p 1, p 2..., p m]; Definition Φ is operating as the capable vector of non-zero corresponding to extraction matrix and is combined into submatrix, the definition mask code matrix
Figure FDA00002811784700013
It is all 1 to HEAD, LWRIST, RWRIST, LHEEL, dimension element that the RHEEL articulation point is relevant, and other row elements are 0; Database feature matrix
Figure FDA00002811784700017
Figure FDA00002811784700014
Wherein k=5 * 3=15 represent control signal and k ?the dimension of d tree, at last to D build k ?d tree index;
Step 5: search in database with step 2 in the athletic posture that mates most of the articulation point control signal of catching: for the control signal of the input of t frame Wherein k is for controlling the number of articulation point, and 3 in 3*k represents x, y, three coordinate dimensions of z; Utilize the k-d tree to search n the proper vector that approaches the most with it in database Define the matrix of this n neighbour's feature composition for recovering dictionary Database attitude submatrix corresponding to these features is
Figure FDA00002811784700022
Step 6: the articulation point information that step 2 is caught is carried out the linear regression based on neighbour's attitude, recovers athletic posture under local coordinate system: because target is to make the attitude that recovers by P tIn linear composition of attitude, find the solution reconstructed coefficients ω so problem is converted into t, make the attitude of recovery
Figure FDA00002811784700023
The coefficient solution formula is the energy minimization equation, is defined as:
ω t ^ = arg min ω t ( E control ( ω t ) + λ 2 E smooth ( ω t ) ) ,
Wherein,
Figure FDA00002811784700025
Be the reconstructed coefficients that to find the solution, E Control(ω t) the expression control item, i.e. the attitude that assurance recovers is wanted the constraint requirements of As soon as possible Promising Policy low-dimensional control signal, E Smooth(ω t) the level and smooth item of expression, the constraint iff considering current control signal may make the frame of recovery and the shake that high frequency appears in interframe, adds constraint by the attitude that former frame is recovered, and can guarantee the flatness requirement, λ 2The weight that expression is level and smooth;
Definition:
E control ( ω t ) | | Ds t * ω t - y t | | 2 2 + λ 1 | | ω t | | 1 ,
Wherein, y tBe current time low-dimensional control signal, Ds tThe control that a small amount of articulation point data of serving as reasons consist of recovers dictionary, λ 1Represent sparse coefficient, when a small amount of articulation point corresponding to the attitude that guarantees to recover approaches as far as possible with control signal, the sparse property of assurance reconstructed coefficients;
Definition:
Figure FDA00002811784700027
Wherein,
Figure FDA00002811784700028
Be the whole body attitude that former frame recovers, the pose recovery dictionary that neighbour's attitude that Pt represents to be indexed by control signal consists of by the former frame attitude, can retrain non-control articulation point; With these two substitution formula, can obtain complete formula:
Figure FDA00002811784700029
Wherein,
Figure FDA000028117847000210
Can merge and obtain:
Figure FDA000028117847000211
The user finds the solution top formula to alternative algorithm, wherein sparse regular terms parameter lambda 1Be set to 0.1, a level and smooth parameter lambda 2Be set to 0.2; The reconstruction coefficients of finding the solution out multiply by the reconstruct dictionary, athletic posture under the local coordinate system that can be restored out;
Step 7: according to the root node information of catching in step 2, the attitude that recovers is carried out translation and rotate obtaining its position under global coordinate system, be the human body attitude that finally recovers.
CN201310047891.1A 2013-02-02 2013-02-02 A kind of human body attitude sparse reconstruction method based on key signature point Active CN103150752B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201310047891.1A CN103150752B (en) 2013-02-02 2013-02-02 A kind of human body attitude sparse reconstruction method based on key signature point

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201310047891.1A CN103150752B (en) 2013-02-02 2013-02-02 A kind of human body attitude sparse reconstruction method based on key signature point

Publications (2)

Publication Number Publication Date
CN103150752A true CN103150752A (en) 2013-06-12
CN103150752B CN103150752B (en) 2015-08-26

Family

ID=48548801

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201310047891.1A Active CN103150752B (en) 2013-02-02 2013-02-02 A kind of human body attitude sparse reconstruction method based on key signature point

Country Status (1)

Country Link
CN (1) CN103150752B (en)

Cited By (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104360743A (en) * 2014-11-20 2015-02-18 姚尧 Body posture data acquisition method and system, and data processing device
CN104463945A (en) * 2013-11-25 2015-03-25 安徽寰智信息科技股份有限公司 Human body posture rebuilding method
CN104715493A (en) * 2015-03-23 2015-06-17 北京工业大学 Moving body posture estimating method
CN105825549A (en) * 2016-03-17 2016-08-03 电子科技大学 Articulated joint point spatial motion reconstruction method based on image sequence
CN106683124A (en) * 2017-01-04 2017-05-17 陈清 Motion data collection device with location stamp and motion data collection system
CN107806837A (en) * 2017-10-29 2018-03-16 北京工业大学 A kind of non-intrusion type wrist joint axis motion model measuring method
CN109032339A (en) * 2018-06-29 2018-12-18 贵州威爱教育科技有限公司 A kind of method and system that real-time intelligent body-sensing is synchronous
CN112037312A (en) * 2020-11-04 2020-12-04 成都市谛视科技有限公司 Real-time human body posture inverse kinematics solving method and device
CN112560959A (en) * 2020-12-17 2021-03-26 咪咕文化科技有限公司 Skeleton animation posture matching method and device, electronic equipment and storage medium
CN114393575A (en) * 2021-12-17 2022-04-26 重庆特斯联智慧科技股份有限公司 Robot control method and system based on high-efficiency recognition of user posture
US11421980B2 (en) * 2017-11-22 2022-08-23 Ford Global Technologies, Llc Method for determining a position and orientation of an object using a profilometer

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8139067B2 (en) * 2006-07-25 2012-03-20 The Board Of Trustees Of The Leland Stanford Junior University Shape completion, animation and marker-less motion capture of people, animals or characters
CN102651072A (en) * 2012-04-06 2012-08-29 浙江大学 Classification method for three-dimensional human motion data

Cited By (19)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104463945A (en) * 2013-11-25 2015-03-25 安徽寰智信息科技股份有限公司 Human body posture rebuilding method
US10395106B2 (en) 2014-11-20 2019-08-27 Wuhan Zhunwofei Science and Technology Co., LTD Human body posture data acquisition method and system, and data processing device
WO2016078131A1 (en) * 2014-11-20 2016-05-26 姚尧 Human body posture data acquisition method and system, and data processing device
CN104360743B (en) * 2014-11-20 2017-05-31 武汉准我飞科技有限公司 The acquisition methods of human body attitude data, system and data processing equipment
CN104360743A (en) * 2014-11-20 2015-02-18 姚尧 Body posture data acquisition method and system, and data processing device
CN104715493A (en) * 2015-03-23 2015-06-17 北京工业大学 Moving body posture estimating method
CN104715493B (en) * 2015-03-23 2018-01-19 北京工业大学 A kind of method of movement human Attitude estimation
CN105825549A (en) * 2016-03-17 2016-08-03 电子科技大学 Articulated joint point spatial motion reconstruction method based on image sequence
CN105825549B (en) * 2016-03-17 2018-09-21 电子科技大学 A kind of articulation joint space of points motor reorgan method towards image sequence
CN106683124A (en) * 2017-01-04 2017-05-17 陈清 Motion data collection device with location stamp and motion data collection system
CN107806837A (en) * 2017-10-29 2018-03-16 北京工业大学 A kind of non-intrusion type wrist joint axis motion model measuring method
US11421980B2 (en) * 2017-11-22 2022-08-23 Ford Global Technologies, Llc Method for determining a position and orientation of an object using a profilometer
CN109032339A (en) * 2018-06-29 2018-12-18 贵州威爱教育科技有限公司 A kind of method and system that real-time intelligent body-sensing is synchronous
CN112037312A (en) * 2020-11-04 2020-12-04 成都市谛视科技有限公司 Real-time human body posture inverse kinematics solving method and device
CN112037312B (en) * 2020-11-04 2021-02-09 成都市谛视科技有限公司 Real-time human body posture inverse kinematics solving method and device
CN112560959A (en) * 2020-12-17 2021-03-26 咪咕文化科技有限公司 Skeleton animation posture matching method and device, electronic equipment and storage medium
CN112560959B (en) * 2020-12-17 2024-06-04 咪咕文化科技有限公司 Gesture matching method and device for bone animation, electronic equipment and storage medium
CN114393575A (en) * 2021-12-17 2022-04-26 重庆特斯联智慧科技股份有限公司 Robot control method and system based on high-efficiency recognition of user posture
CN114393575B (en) * 2021-12-17 2024-04-02 重庆特斯联智慧科技股份有限公司 Robot control method and system based on high-efficiency recognition of user gestures

Also Published As

Publication number Publication date
CN103150752B (en) 2015-08-26

Similar Documents

Publication Publication Date Title
CN103150752B (en) A kind of human body attitude sparse reconstruction method based on key signature point
Xu et al. Mo 2 cap 2: Real-time mobile 3d motion capture with a cap-mounted fisheye camera
CN111460875B (en) Image processing method and apparatus, image device, and storage medium
Rhodin et al. Egocap: egocentric marker-less motion capture with two fisheye cameras
Xia et al. A survey on human performance capture and animation
Wei et al. Videomocap: Modeling physically realistic human motion from monocular video sequences
Zhao et al. Combining marker-based mocap and rgb-d camera for acquiring high-fidelity hand motion data
Andrews et al. Real-time physics-based motion capture with sparse sensors
Chen et al. KinÊtre: animating the world with the human body
CN113706699B (en) Data processing method and device, electronic equipment and computer readable storage medium
CN110033505A (en) A kind of human action capture based on deep learning and virtual animation producing method
JP2023502795A (en) A real-time system for generating 4D spatio-temporal models of real-world environments
CN110140147A (en) Video frame synthesis with deep learning
CN110188700B (en) Human body three-dimensional joint point prediction method based on grouping regression model
Ren et al. Lidar-aid inertial poser: Large-scale human motion capture by sparse inertial and lidar sensors
Uddin et al. Human Activity Recognition via 3-D joint angle features and Hidden Markov models
CN102289836A (en) Method for synthesizing plant animation
JP2002269580A (en) Animation creating system
Huang et al. A review of 3D human body pose estimation and mesh recovery
Tao et al. Human modeling and real-time motion reconstruction for micro-sensor motion capture
CN102930516B (en) Data driven and sparsely represented three-dimensional human motion denoising method
US10410393B2 (en) Animating a virtual object in a virtual world
Jiang et al. EgoPoser: Robust Real-Time Ego-Body Pose Estimation in Large Scenes
TWI534756B (en) Motion-coded image, producing module, image processing module and motion displaying module
Kobayashi et al. Motion capture dataset for practical use of AI-based motion editing and stylization

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C14 Grant of patent or utility model
GR01 Patent grant