CN1975779A - Three-dimensional human body movement data dividing method - Google Patents
Three-dimensional human body movement data dividing method Download PDFInfo
- Publication number
- CN1975779A CN1975779A CNA2006100533919A CN200610053391A CN1975779A CN 1975779 A CN1975779 A CN 1975779A CN A2006100533919 A CNA2006100533919 A CN A2006100533919A CN 200610053391 A CN200610053391 A CN 200610053391A CN 1975779 A CN1975779 A CN 1975779A
- Authority
- CN
- China
- Prior art keywords
- data
- frame
- human body
- sequence
- motion
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 47
- 238000004458 analytical method Methods 0.000 claims abstract description 7
- 239000011159 matrix material Substances 0.000 claims description 11
- 238000006073 displacement reaction Methods 0.000 claims description 10
- 241001269238 Data Species 0.000 claims description 8
- 238000000638 solvent extraction Methods 0.000 claims description 7
- 230000007704 transition Effects 0.000 claims description 6
- 238000001514 detection method Methods 0.000 claims description 3
- 238000005192 partition Methods 0.000 abstract 3
- 230000006399 behavior Effects 0.000 description 12
- 230000003287 optical effect Effects 0.000 description 5
- 230000011218 segmentation Effects 0.000 description 5
- 238000006243 chemical reaction Methods 0.000 description 3
- 102100028524 Lysosomal protective protein Human genes 0.000 description 2
- 101710162021 Lysosomal protective protein Proteins 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 210000000988 bone and bone Anatomy 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 239000012634 fragment Substances 0.000 description 1
- 230000008676 import Effects 0.000 description 1
- 230000009191 jumping Effects 0.000 description 1
- 238000013179 statistical model Methods 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
Images
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
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Psychiatry (AREA)
- Social Psychology (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Processing Or Creating Images (AREA)
Abstract
The invention discloses a partition method on human motion 3D data. First, use manifold analysis method ISOMAP to map the original human sequence motion data to the low dimensional manifold space; secondly, utilize a heuristic method to detect rough partition points of different types of movement in the motion data sequence, then long sequence of human motion data are partitioned to several segments; last, for every two segments concatenated in time in the low dimensional manifold space use K-average clustering algorithm to compute the accurate partition point.
Description
Technical field
The present invention relates to computer three-dimensional animation technology and multi-medium data process field, relate in particular to a kind of three-dimensional human body movement data dividing method.
Background technology
Along with being extensive use of of capturing movement equipment, generate a large amount of realistic 3 d human motion datas in recent years, these data are widely used in computer game, animation generates and a plurality of fields such as medical science emulation.Because the human motion sequence that adopts optical acquisition system to obtain when going into the body capturing movement is longer, and often comprised plurality of continuous but have dissimilar exercise datas (such as walk, run, jumping etc.), store, retrieve, browse and motion editing further for the ease of the compression of a large amount of 3 d human motion datas, is very useful to growing that the dissimilar exercise data that comprises in the sequence human body movement data cuts apart automatically.
It is a kind of technology that is widely used in the multi-medium data analysis field that data are cut apart, and cuts apart audio segmentation etc. such as video lens.Cut apart differently with other multi-medium data, 3 d human motion data is a kind of structurized multi-medium data.Existing human body movement data dividing method mainly is divided into two classes: based on the exercise data dividing method and the motion segmentation method of analyzing based on bottom data of model." american computer association graphics journal " be (ACM Transactions on Graphics2003 in the third phase in 2003,33 (3): 402-408) announced a kind of motion mask method based on model, based on the training sample set of manual mark and the support vector machine classifier that builds, the different motion type in the exercise data sequence can be marked automatically.(Proceedings of ICIP 2003 in the international Flame Image Process conference collections of thesis in 2003, II:105-108) announced a kind of human body behavior dividing method of layering, adopt a kind of dynamic layered data structure table body exercise data of leting others have a look at, and then adopt a kind of simple Bayes classifier to finish the human body behavior to cut apart.Need be based on the exercise data dividing method of model based on a large amount of training datas, and segmentation result to be subjected to the training data influence bigger, therefore be applied in practice than difficulty.
(Autonomous Robots in " autonomous robot journal " first phase in 2002,2002,12 (1): method 39-54) adopts the method that detects joint angles data zero crossing that the human arm exercise data is cut apart, and this method realizes very simple, but the segmentation result precision is not high.Graphic interface conference meeting paper was concentrated (Proceedings of GraphicsInterface in 2005,2004,185-194) method of Gong Buing thinks that dissimilar human body movement datas should have different inner dimensions, therefore adopt PCA, PPCA, methods such as GMM are carried out the dimension analysis to the human body movement data of long sequence, and then reach the purpose that data are cut apart, this method can reach higher cutting precision, and do not need the training data support, but PCA, setting up of statistical models such as the structure of PPCA subspace and GMM is consuming time too much, and it is not high therefore to cut apart efficient.
Summary of the invention
The present invention is the shortcoming and defect that overcomes above-mentioned existing method, and a kind of efficient, automatic three-dimensional human body movement data dividing method is provided.
Three-dimensional human body movement data dividing method may further comprise the steps:
(1) at first adopts skeleton model representation 3 d human motion data, and when pre-service, will fall with root node displacement and the rotating vector data filter that the human motion type has nothing to do with 16 articulation points;
(2) the rough partitioning algorithm of the heuristic human body movement data of employing is determined the rough cut-point in the whole human body exercise data sequence;
(3) be the boundary with rough cut-point, original human body movement data sequence is divided into plurality of sections, per two sections exercise datas as input, are adopted and determine its accurate cut-point based on three-dimensional motion data Nonlinear Dimension Reduction and the dividing method of ISOMAP.
Described 3 d human motion data method for expressing is that human body is expressed as the tree shaped model that is made of 16 articulation points, and represents 3 d human motion data in the following way:
M={F(1),F(2),...,F(t),...,F(n)}
F (t)={ p (t), q
1(t) ..., q
m(t) } wherein F (t) is a t frame among the three-dimensional motion sequence M, and p (t) is the displacement of root node in the t frame, q
i(t) be the rotating vector of i articulation point in the t frame.In addition, when the 3 d human motion data sequence was calculated, the displacement of human body root node and rotating vector data can be filtered.
The rough partitioning algorithm of described heuristic human body movement data is that the distance of the Euler between each frame in the whole 3 d human motion sequence of input is calculated, the position that forms the distance matrix of whole motion sequence and be chosen in first frame matrix section of adjusting the distance, and then adopt a kind of didactic digital signal transition detection method just can obtain the rough cut-point of whole motion sequence, the false code of this heuristic is as follows:
M=LoadMotion (...); ∥ loads exercise data
D=DistanceMatrix (D); ∥ calculates the distance matrix of motion sequence
Len=Length (M); ∥ obtains motion sequence length
CurFrm=1; It is first frame that ∥ is provided with present frame
while(curFrm<len)
{
CurDistCurv=D (curFrm :); ∥ obtains the distance Curve of present frame, and sectioning
Set interval; ∥ is provided with the algorithm step-length
∥ calculates the ultimate range between the frame and frame in the current motion sequence (from the curFrm+sp frame to the curFrm+ep frame)
diff=MaxMinDiff(curDistCurv(curFrm+sp:curFrm+ep));
i=curFrm+ep;
while(i<len)
{
[maxvalue?maxindex]=max(curDistCurv(i:i+interval));
[minvalue?minindex]=min(curDistCurv(i:i+interval));
if(maxvalue-minvalue)>=diff*a
{
segPt=i+fix((maxindex+minindex)/2);
break;
}
if(maxvalue-minvalue)<=diff*β
{
segPt=i+fix((maxindex+miniudex)/2);
break;
}
i=i+interval;
}
Save(segPt);
curFrm=segPt;
}
Described three-dimensional motion data Nonlinear Dimension Reduction and dividing method based on ISOMAP is as a higher-dimension sample point with each frame in the 3 d human motion sequence, whole 3 d human motion sequence is as sample set, be input to and carry out computing in the ISOMAP algorithm, the 3 d human motion sequence samples point that output obtains in the low-dimensional stream shape space distributes; Adopt the average clustering algorithm of K that the sample point that distributes in these low-dimensional stream shape spaces is carried out cluster then, obtain final accurate cut-point.
Three-dimensional human body movement data dividing method of the present invention has following characteristics:
1. this method computation complexity is low, and is insensitive to the data dimension based on the Nonlinear Dimension Reduction algorithm of stream conformal analysis ISOMAP, so counting yield is higher;
2. this method automaticity height, given one section comprise polytype human body movement data after, do not need to want the set-up and calculated parameter to calculate automatically;
3. the data segmentation result accuracy height that obtains based on the Nonlinear Dimension Reduction algorithm.
Description of drawings
Fig. 1 is the human skeleton model;
Fig. 2 represents the distance matrix of 3 d human motion sequence, and red dotted line represents it may is a cut-point here;
Fig. 3 is the section expression that distance matrix goes out at first frame, and red little frame table shows it may is a cut-point here;
Fig. 4 is a three-dimensional human body movement data dividing method workflow diagram of the present invention;
Fig. 5 is a human body movement data through the ISOMAP Nonlinear Dimension Reduction and the result after cutting apart, and wherein (a) is the distribution of 3 d human motion data in the three-dimensional manifold space, (b) is the result who adopts the K average algorithm that this motion sequence is cut apart;
Fig. 6 is the result that one section long sequence 3 d human motion data that comprises various human body behaviors is cut apart.
Embodiment
Concrete technical scheme and the step of implementing of three-dimensional human body movement data dividing method of the present invention is as follows:
1. 3 d human motion data pre-service
The present invention adopts human skeleton model as shown in Figure 1.This model is made up of 16 joints.Exercise data can be expressed as form:
M={F(1),F(2),...,F(t),...,F(n)} (1)
F (t)={ p (t), q
1(t) ..., q
m(t) } (2) wherein F (t) be t frame among the three-dimensional motion sequence M, p (t) is the displacement of root node in the t frame, q
i(t) be the rotating vector of i articulation point in the t frame.
When the 3 d human motion data sequence is calculated, the displacement of human body root node and rotating vector data can be filtered, because the displacement of root node and rotating vector data only represent this moment human body the locus and towards, it doesn't matter with human body behavior itself.
2. heuristic rough partitioning algorithm
After obtaining previously described 3 d human motion data and representing, need cut apart roughly it.Here the distance of the Euler between each frame in the whole 3 d human motion sequence of input is calculated, form the distance matrix (seeing accompanying drawing 2) of whole motion sequence, can see this section 3 d human motion sequence roughly cutting be 4 or 5 sections (seeing the position of red dotted line in the accompanying drawing 2).
But want directly to find out very difficulty of cut-point with computer approach from Fig. 2, the position that therefore is chosen in first frame matrix section of adjusting the distance obtains the figure shown in the accompanying drawing 3.As seen from Figure 3, adopt a kind of didactic digital signal transition detection method just can obtain the rough cut-point of whole motion sequence.
The heuristic rough dividing method false code of motion sequence is as follows:
M=LoadMotion (...); ∥ loads exercise data
D=DistanceMatrix (D); ∥ calculates the distance matrix of motion sequence
Len=Length (M); ∥ obtains motion sequence length
CurFrm=1; It is first frame that ∥ is provided with present frame
while(curFrm<len)
{
CurDistCurv=D (curFrm :); ∥ obtains the distance Curve of present frame, and sectioning
Set interval; ∥ is provided with the algorithm step-length
∥ calculates the ultimate range between the frame and frame in the current motion sequence (from the curFrm+sp frame to the curFrm+ep frame)
diff=MaxMinDiff(curDistCurv(curFrm+sp:curFrm+ep));
i=curFrm+ep;
while(i<len)
{
[maxvalue?maxindex]=max(curDistCurv(i:i+interval));
[minvalue?minindex]=min(curDistCurv(i:i+interval));
if(maxvalue-minvalue)>=diff*a
{
segPt=i+fix((maxindex+minindex)/2);
break;
}
if(maxvalue-minvalue)<=diff*β
{
segPt=i+fix((maxindex+minindex)/2);
break;
}
i=i+interval;
}
Save(segPt);
curFrm=segPt;}
3. based on the three-dimensional motion data Nonlinear Dimension Reduction of ISOMAP and cut apart
Because 3 d human motion data dimension height, and in the middle of original data space complex distribution, even high distortion or folding.Even therefore after obtaining the rough cut-point of whole 3 d human motion sequence, also be difficult in central definite its accurate cut-point of original data space.
Consider the 3 d human motion data sequence be a kind of in the space branch's high complexity, nonlinear data, here adopt Nonlinear Dimension Reduction algorithm that it is carried out dimensionality reduction, higher-dimension 3 d human motion data sequence is mapped in the simple relatively low-dimensional stream of the structure shape space handles based on ISOMAP.ISOMAP is a kind of comparatively ripe Nonlinear Dimension Reduction method, in the present invention with each frame in the 3 d human motion sequence as a higher-dimension sample point, whole 3 d human motion sequence is input to and carries out computing in the ISOMAP algorithm as sample set.After obtaining the distribution of 3 d human motion data in the low-dimensional popular world, adopt the average clustering algorithm of K that sample point is carried out cluster and can obtain accurate 3 d human motion data cut-point.
Accompanying drawing 4 shows human body movement data automatic division method workflow diagram of the present invention.The concrete implementing procedure of this method comprise 3 d human motion data input 10, data pre-service 20, motion sequence cut apart roughly 30 and motion sequence accurately cut apart 40.
3 d human motion data input 10, the 3 d human motion data here comprises by optical motion capture device and relevant speciality software (as Maya, Motion Builder etc.) 3 d human motion data of the various forms of Chan Shenging, the capture device MotionAnalysis Hawk that adopts U.S. Motion Analysis company to produce as this example gathers various types of 3 d human motion datas.
Data pre-service 20,3 d human motion data to the various forms of user input carries out format conversion, the bone of different topology structure is converted to the manikin that uses among the present invention, 3 d human motion data is converted to the form of the present invention definition and root node displacement and rotating vector data filter are fallen.
Motion sequence cuts apart 30 roughly, go out to read in the 3 d human motion data sequence of handling well from data pre-service 20, adopt heuristic rough partitioning algorithm to determine the rough cut-point that exists in the whole human body motion sequence, original human body motion sequence is divided into plurality of sections.
Motion sequence accurately cuts apart 40, cut apart the human motion sequence fragment that reads in 30 after two sections processes are cut apart roughly roughly from motion sequence at every turn, adopt the ISOMAP algorithm that it is carried out being mapped to low-dimensional stream shape space after the Nonlinear Dimension Reduction, and adopt the average clustering algorithm of K that the sample point in the low-dimensional stream shape space that obtains is carried out cluster, finally obtain the accurate cut-point between two sections motion sequence segments.
As shown in Figure 5, provided the example that one section three-dimensional motion sequence that comprises two kinds of human body behaviors is carried out the ISOMAP dimensionality reduction and accurately cut apart.Describe the concrete steps that this example is implemented in detail below in conjunction with method of the present invention, as follows:
(1) obtains one section 3 D human body animation sequence that comes from optical motion capture systems or the generation of professional animation soft, data in this example come from optical motion capture systems (TRC data layout), comprised normal walking and the two kinds of behaviors that go sideways, and be the nature transition between them;
(2), adopt existing exercise data conversion method that the TRC data are converted to the spin data presentation format with 16 articulation points that satisfies the present invention's definition and will represent that the translation of root node and spin data filter out to catch TRC form the original motion data of obtaining in the step (1) as input;
(3) because the motion sequence in this example only includes two kinds of forms of motion, therefore not needing to carry out motion sequence cuts apart roughly, carry out Nonlinear Dimension Reduction but whole section exercise data inputed to the ISOMAP algorithm, obtain as Fig. 5 (a) the distribution of the low-dimensional that is shown on popular;
(4) with the data sample point on the low-dimensional that obtains in the step (3) the stream shape as importing, the average clustering algorithm of employing K can obtain the accurate cut-point of final exercise data, shown in Fig. 5 (b).
The result of this example shows in accompanying drawing 5, can see by the Nonlinear Dimension Reduction of ISOMAP and handling and the average clustering algorithm of K, can very easily find the solution two accurate cut-points between the natural transition behavior.
Embodiment 2
As shown in Figure 6, provided the result that one section three-dimensional motion sequence that comprises multiple human body behavior is cut apart.Describe the concrete steps that this example is implemented in detail below in conjunction with method of the present invention, as follows:
(1) this example adopts the optical motion capture systems to obtain the original 3 d human motion sequence of one section TRC form, comprise multiple behavior, be successively walk, lean to one side away, mop the floor, walk, squat down beat ground, the 8 kinds of behaviors such as clean ground, the cleaned window of standing, walk of squatting down, and be the nature transition between each behavior;
(2), adopt existing exercise data conversion method that the TRC data are converted to the spin data presentation format with 16 articulation points that satisfies the present invention's definition and will represent that the translation of root node and spin data filter out to catch TRC form the original motion data of obtaining in the step (1) as input;
(3) the long sequence human body movement data that uses heuristic rough partitioning algorithm of the present invention to import is divided into 8 sections roughly, is labeled as S respectively
1... S
i... S
8, at initialization i=1 in season;
(4) with S
iWith S
I+1Be input in the ISOMAP algorithm kind as the continuous motion sequence and carry out Nonlinear Dimension Reduction, obtain its distribution in low-dimensional stream shape space;
(5) with the data sample point on the low-dimensional that obtains in the step (4) the stream shape as importing, the average clustering algorithm of employing K can obtain S
iWith S
I+1Between accurate cut-point;
(6) make i=i+1, get back to step (4), up to having found the solution each accurate cut-point.
The result of this example shows in accompanying drawing 6, can see by to cut apart roughly that per two the adjacent data segments that obtain carry out that the ISOMAP Nonlinear Dimension Reduction is handled and the average cluster of K after, can accurately obtain the cut-point between each behavior.
Claims (4)
1. three-dimensional human body movement data dividing method is characterized in that may further comprise the steps:
(1) at first adopts skeleton model representation 3 d human motion data, and when pre-service, will fall with root node displacement and the rotating vector data filter that the human motion type has nothing to do with 16 articulation points;
(2) the rough partitioning algorithm of the heuristic human body movement data of employing is determined the rough cut-point in the whole human body exercise data sequence;
(3) be the boundary with rough cut-point, original human body movement data sequence is divided into plurality of sections, per two sections exercise datas as input, are adopted based on three-dimensional motion data Nonlinear Dimension Reduction and the dividing method of stream conformal analysis ISOMAP and determine its accurate cut-point.
2. according to a kind of three-dimensional human body movement data dividing method described in the claim 1, it is characterized in that: described 3 d human motion data method for expressing is that human body is expressed as the tree shaped model that is made of 16 articulation points, and represents 3 d human motion data in the following way:
M={F(1),F(2),...,F(t),...,F(n)}
F(t)={p(t),q
1(t),...,q
m(t)}
Wherein F (t) is the t frame among the three-dimensional motion sequence M, and p (t) is the displacement of root node in the t frame, q
i(t) be the rotating vector of i articulation point in the t frame, in addition, when the 3 d human motion data sequence was calculated, the displacement of human body root node and rotating vector data can be filtered.
3. according to a kind of three-dimensional human body movement data dividing method described in the claim 1, it is characterized in that: the rough partitioning algorithm of described heuristic human body movement data is that the distance of the Euler between each frame in the whole 3 d human motion sequence of input is calculated, the position that forms the distance matrix of whole motion sequence and be chosen in first frame matrix section of adjusting the distance, and then adopt a kind of didactic digital signal transition detection method just can obtain the rough cut-point of whole motion sequence, the false code of this heuristic is as follows:
M=LoadMotion (...); // loading exercise data D=DistanceMatrix (D); The distance matrix len=Length (M) of // calculating motion sequence; // obtain motion sequence length curFrm=1; // present frame be set be the first frame the while ({ curDistCurv=D (curFrm :) of curFrm<len); // obtain the distance Curve of present frame, reach sectioning Set interval; // ultimate range between the frame and frame in algorithm step-length // current motion sequence of calculating (from the curFrm+sp frame to the curFrm+ep frame) diff=MaxMinDiff (curDistCurv (curFrm+sp:curFrm+ep)) is set; I=curFrm+ep; While ({ [maxvalue the maxindex]=max (curDistCurv (i:i+interval)) of i<len); [minvalue minindex]=min (curDistCurv (i:i+interval)); If (maxvalue-minvalue)>=diff* α { segPt=i+fix ((maxindex+minindex)/2); Break; If (maxvalue-minvalue)<=diff* β { segPt=i+fix ((maxindex+minindex)/2); Break; I=i+interval; Save (segPt); CurFrm=segPt; }
。
4. according to a kind of three-dimensional human body movement data dividing method described in the claim 1, it is characterized in that: described three-dimensional motion data Nonlinear Dimension Reduction and dividing method based on stream conformal analysis ISOMAP is as a higher-dimension sample point with each frame in the 3 d human motion sequence, whole 3 d human motion sequence is as sample set, be input in the stream conformal analysis ISOMAP algorithm and carry out computing, the 3 d human motion sequence samples point that output obtains in the low-dimensional stream shape space distributes; Adopt the average clustering algorithm of K that the sample point that distributes in these low-dimensional stream shape spaces is carried out cluster then, obtain final accurate cut-point.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CNA2006100533919A CN1975779A (en) | 2006-09-14 | 2006-09-14 | Three-dimensional human body movement data dividing method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CNA2006100533919A CN1975779A (en) | 2006-09-14 | 2006-09-14 | Three-dimensional human body movement data dividing method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN1975779A true CN1975779A (en) | 2007-06-06 |
Family
ID=38125825
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CNA2006100533919A Pending CN1975779A (en) | 2006-09-14 | 2006-09-14 | Three-dimensional human body movement data dividing method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN1975779A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101276370B (en) * | 2008-01-14 | 2010-10-13 | 浙江大学 | Three-dimensional human body movement data retrieval method based on key frame |
CN101866494A (en) * | 2010-06-28 | 2010-10-20 | 北京理工大学 | Method for carrying out segmentation on role model by utilizing grid vertexes |
CN102521843A (en) * | 2011-11-28 | 2012-06-27 | 大连大学 | Three-dimensional human body motion analysis and synthesis method based on manifold learning |
CN106066996A (en) * | 2016-05-27 | 2016-11-02 | 上海理工大学 | The local feature method for expressing of human action and in the application of Activity recognition |
CN106580338A (en) * | 2016-11-15 | 2017-04-26 | 南方医科大学 | Maximum length sequence optimization method and system for nonlinear system identification |
-
2006
- 2006-09-14 CN CNA2006100533919A patent/CN1975779A/en active Pending
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101276370B (en) * | 2008-01-14 | 2010-10-13 | 浙江大学 | Three-dimensional human body movement data retrieval method based on key frame |
CN101866494A (en) * | 2010-06-28 | 2010-10-20 | 北京理工大学 | Method for carrying out segmentation on role model by utilizing grid vertexes |
CN101866494B (en) * | 2010-06-28 | 2012-06-27 | 北京理工大学 | Method for carrying out segmentation on role model by utilizing grid vertexes |
CN102521843A (en) * | 2011-11-28 | 2012-06-27 | 大连大学 | Three-dimensional human body motion analysis and synthesis method based on manifold learning |
CN102521843B (en) * | 2011-11-28 | 2014-06-04 | 大连大学 | Three-dimensional human body motion analysis and synthesis method based on manifold learning |
CN106066996A (en) * | 2016-05-27 | 2016-11-02 | 上海理工大学 | The local feature method for expressing of human action and in the application of Activity recognition |
CN106066996B (en) * | 2016-05-27 | 2019-07-30 | 上海理工大学 | The local feature representation method of human action and its application in Activity recognition |
CN106580338A (en) * | 2016-11-15 | 2017-04-26 | 南方医科大学 | Maximum length sequence optimization method and system for nonlinear system identification |
CN106580338B (en) * | 2016-11-15 | 2020-02-21 | 南方医科大学 | Maximum length sequence optimization method and system for nonlinear system identification |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN1138231C (en) | Image generation apparatus, its method, game machine and recording medium | |
CN1967525A (en) | Extraction method of key frame of 3d human motion data | |
US20140285496A1 (en) | Data Compression for Real-Time Streaming of Deformable 3D Models for 3D Animation | |
CN1279761C (en) | Image processing apparatus, image processing method, record medium, computer program, and semiconductor device | |
CN100343874C (en) | Voice-based colored human face synthesizing method and system, coloring method and apparatus | |
CN1975779A (en) | Three-dimensional human body movement data dividing method | |
CN1734503A (en) | Stretch-driven mesh parameterization method using spectral analysis | |
CN1799069A (en) | Statistical dynamic collisions method and apparatus | |
CN1444185A (en) | Computer image animation device and method | |
CN1920886A (en) | Video flow based three-dimensional dynamic human face expression model construction method | |
EP1321894A3 (en) | Apparatus and method for representing 3-dimensional objects unsing depth images | |
CN1130667C (en) | Device and method for controlling quality of reproduction of motion picture | |
CN101051390A (en) | Method and system for simulating river | |
CN1831468A (en) | Method for deciding relative position of laser scanner and robot | |
JP5109191B2 (en) | FACS (Facial Expression Coding System) cleaning in motion capture | |
CN101046891A (en) | Texture mapping apparatus and method | |
CN101038678A (en) | Smooth symmetrical surface rebuilding method based on single image | |
CN1873647A (en) | CAD method, CAD system and program storage medium storing CAD program thereof | |
CN1198684A (en) | Video game machine and medium where video game program is stored | |
CN1932885A (en) | Three-dimensional image processing | |
CN101533528A (en) | Optical motion capture data processing method based on module piecewise linear model | |
CN101069214A (en) | System and method for tracking facial muscle and eye motion for computer graphics animation | |
CN1695151A (en) | CAD system and CAD program | |
CN107730501A (en) | A kind of fish body tail frequency detection method and system | |
CN1656518A (en) | Three dimensional image processing program, three dimensional image processing method, and video game device |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
WD01 | Invention patent application deemed withdrawn after publication |