CN101308500A - Visual and efficient three-dimensional human body movement data retrieval method based on demonstrated performance - Google Patents
Visual and efficient three-dimensional human body movement data retrieval method based on demonstrated performance Download PDFInfo
- Publication number
- CN101308500A CN101308500A CNA2008100616728A CN200810061672A CN101308500A CN 101308500 A CN101308500 A CN 101308500A CN A2008100616728 A CNA2008100616728 A CN A2008100616728A CN 200810061672 A CN200810061672 A CN 200810061672A CN 101308500 A CN101308500 A CN 101308500A
- Authority
- CN
- China
- Prior art keywords
- motion
- dimensional human
- characteristic
- retrieval
- motion sensor
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Landscapes
- Processing Or Creating Images (AREA)
Abstract
The invention discloses an intuitive and efficient three-dimensional human motion data retrieval method based on demonstration performance. The method comprises the following steps: first, to constitute a motion index for a large-capacity motion database through subspace partition; second, to set an overall reference coordinate system of a motion sensor; third, to load a standard three-dimensional human skeleton; fourth, to designate the corresponding relationship between the motion sensor and the standard three-dimensional human skeleton nodes; fifth, to bind the motion sensor and the standard three-dimensional human skeleton through posture alignment; sixth, to drive the standard three-dimensional human skeleton through the data acquired by the motion sensor to get demonstration performance motion; seventh, to make feature extraction to the demonstration performance motion through subspace partition; eighth, to load the motion index and carry out qualitative retrieval; ninth, to set parameters, and then to carry out quantitative retrieval based on the qualitative retrieval. The invention solves the problem of how to intuitively and accurately express the creation intention of animation producers in the process of making character animation to achieve rapid and efficient extraction of required motion data.
Description
Technical field
The present invention relates to the search method of 3 d human motion data, especially a kind of intuitive and efficient three-dimensional human body movement data retrieval method based on demonstration performance.
Background technology
According to the difference of input information, motion retrieval mainly contains four kinds of methods: text, note spectrum language, cartographical sketching, motion sample.Traditional text searching method is described as input (as " kicking after going out fist earlier ") with key word, search out motion segment with these mark attributes, but the textual description content is fuzzy, imperfect, and has the inconsistent problem of subjective understanding, is unsuitable for retrieving jumbo motion database.Remember that the spectrum language lays particular emphasis on the description of the bottom position and the direction of each articulation point, can express the content and the implication [1] of exercise data more clearlyly, but this method needs the cartoon making personnel to be familiar with note spectrum language, directly perceived inadequately aspect alternately.Also can show personage's motion with two-dimension human body skeleton diagram and Freehandhand-drawing people bigraph, can recover three-dimensional skeleton posture according to these cartographical sketchings, carry out motion retrieval [2] with the skeleton posture as input then, but recovering three-dimensional posture from two-dimentional sketch is a underconstrained problem, may reconstruct too many action, add the ambiguity on the match retrieval, may retrieve a large amount of incoherent exercise datas.Content-based motion retrieval is the matching problem of a time series data in essence: as input, find out all motion segments [3,4,5] similarly in the motion database with one section motion sample, this method is one of main stream approach of exercise data retrieval.Wherein the most representative work comes from people such as Muller, they have proposed the qualitative features of the geometric relationship between the different body parts under a certain posture of a series of descriptions, change merging by local space time and obtain adaptive segmentation, carry out motion retrieval [6] efficiently on this basis exercise data.The shortcoming of this method is: the user may not have suitable sample at hand as input.
Content-based motion retrieval method mainly comprises three gordian techniquies: similarity calculating, index construct, dynamic time adjustment.
The similarity standard is the criterion of tolerance motion segment otherness, three kinds of basic kinematic similitude types is arranged: similar on the mathematics, similar, semantically similar in logic.Similarly on the mathematics obtain by calculating two distance functions between the motion, the input data can be the positions, towards, [7,8] such as speed and acceleration.But, only similarly on mathematics may cause incomplete even wrong result for retrieval, because similar in logic motion might not similar on mathematics [3] in people's eye.The similar in logic space-time unchangeability that is generally defined as movement node and bone, such as, the right hand whether in the place ahead of trunk.People such as Muller have introduced the qualitative geometric properties that can obtain the space-time unchangeability, have overcome similar and dissimilar in logic problem [6] on the mathematics.Similar on the semantic level is a more scabrous problem, because it is related to the high level of motion implication understood.Some researcher solves matter of semantics [1,9] with movement mark language such as Labanotation, but up to now, problem also is far from being resolved.Most of searching system still needs the user to come the correct implication of mutual affirmation matching result.
The purpose that makes up the motion index is in order to accelerate retrieval rate, and the large-capacity data storehouse needs the index strategy.People such as Liu adopt the dynamic clustering algorithm based on nearest neighbouring rule to make up an index tree based on the motion hierarchy description, and their motion index tree decides subclass [10] .Li that comprises the motion the most similar with importing sample and Prabhakaran to provide a kind of index as a sorter the tree structure of the exercise data of a series of attributes [11].People such as Muller have proposed a kind of index structuring method [6] of the more perfect qualitative geometric properties of use, and these geometric properties are represented with a series of Boolean variable of geometric relationship between the different body parts of representing under a certain posture.Lin has also proposed a kind of similar index scheme [12], thereby he extracts the expression exercise data that feature can be succinct by a pretreated process.
Two similar in logic exercise datas may show very big change in time and space.Therefore the comparison between the exercise data should be considered the difference on the time shaft, such as, the frame corresponding relation between the exercise data should according to one continuous, the monotonically increasing mapping relations are adjusted [3].Dynamic time adjustment (DTW) can be found the optimum corresponding relation between two time serieses, often is used to determine Time Series Similarity, classification and corresponding region [13].
Motion sensor be a kind of can catch many-sided kinetic characteristic (as acceleration of motion and towards etc.) novel device, along with micro electronmechanical (MEMS) development of technology, its volume and price reduce greatly, can be used for fields such as man-machine interaction and capturing movement.Although it is also relatively more difficult to use motion sensor to obtain high-quality exercise data, can obtain rough demonstration campaign easily, thus the intuitive and accurate creation intention that presents the cartoon making personnel.Therefore, we have proposed the intuitive and efficient three-dimensional human body movement data retrieval method based on demonstration performance.This method has solved that cartoon making personnel input information when using the magnanimity motion database to carry out motion retrieval is difficult to describe, the directly perceived and low difficult problem of recall precision inadequately.
List of references
[1]Yu，T.，Shen，X.，Li，Q.，Geng，W.：Motion?retrieval?based?on?movement?notationlanguage.Computer?Animation?and?Virtual?Worlds.2005，16(3-4)，273-282
[2]Li，Q.，Geng，W，et?al.：MotionMaster：authoring?and?choreographing?Kung-fumotions?by?sketch?drawings.In：SCA’06：Proceedings?of?the?2006?ACMSIGGRAPH/Eurographics?symposium?on?Computer?animation：233-241.
[3]Kovar，L.，Gleicher，M.：Automated?extraction?and?parameterization?of?motions?inlarge?data?sets.ACM?Trans.Graph.2004，23(3)，559-568
[4]Sakamoto，Y.，Kuriyama，S.，Kaneko，T.：Motion?map：image-based?retrieval?andsegmentation?of?motion?data.In：SCA’04：Proceedings?of?the?2004?ACMSIGGRAPH/Eurographics?symposium?on?Computer?animation，259-266.Eurographics?Association，Aire-la-Ville，Switzerland，Switzerland
[5]Ren，L.，Patrick，A.，Efros，A.A.，Hodgins，J.K.，Rehg，J.M.：A?data-drivenapproach?to?quantifying?natural?human?motion.ACM?Trans.Graph.2005，24(3)，1090-1097
[6]Muller，M.，Roder，T.，Clausen，M.：Efficient?contentbased?retrieval?of?motioncapture?data.ACM?Trans.Graph.2005，24(3)，677-685
[7]Arikan，O.，Forsyth，D.A.：Interactive?motion?generation?from?examples.ACMTrans.Graph.2002，21(3)，483-490
[8]Lee，J.，Chai，J.，Reitsma，P.S.A.，Hodgins，J.K.，Pollard，N.S.：Interactive?controlof?avatars?animated?with?human?motion?data.In：SIGGRAPH’02：Proceedingsof?the?29th?annual?conference?on?Computer?graphics?and?interactive?techniques，2002，pp.491-500.ACM，New?York，NY，USA
[9]Chiu，C.Y.，Chao，S.P.，Wu，M.Y.，Yang，S.N.，Lin，H.C.：Content-based?retrievalfor?human?motion?data.Journal?of?Visual?Communication?and?ImageRepresentation，Special?Issue?on?Multimedia?Database?Management?Systems.2004，15(3)，446-466
[10]Liu，F.，Zhuang，Y.，Wu，F.，Pan，Y.：3d?motion?retrieval?with?motion?index?tree.Comput.Vis.Image?Underst.2003，92(2-3)，265-284
[11]Li，C.，Prabhakaran，B.：Indexing?of?motion?capture?data?for?efficient?and?fastsimilarity?search.Journal?of?Computers.2006，Vol.1(3)，pp.35-42
[12]Lin，Y.：Efficient?human?motion?retrieval?in?large?databases.In：GRAPHITE’06：Proceedings?of?the?4th?international?conference?on?Computer?graphics?andinteractive?techniques?in?Australasia?and?Southeast?Asia，2006，pp.31-37.ACMPress，New?York，NY，USA
[13]Salvador，S.，Chan，P.：Toward?accurate?dynamic?time?warping?in?linear?time?andspace.Intelligent?Data?Analysis.2007，11，561-580
Summary of the invention
The purpose of this invention is to provide a kind of intuitive and efficient three-dimensional human body movement data retrieval method based on demonstration performance.
Comprise the steps:
1) method of dividing with the subspace makes up the motion index for the high capacity motion database, each exercise data in the motion database is carried out feature extraction and segmentation, and be converted into subspace characteristic symbol sequence;
2) connect motion sensor, set the overall reference frame of motion sensor;
3) standard three-dimensional human skeleton of input consumer premise justice;
4) mutual or come given step 2 by program interface by the input configuration file) in corresponding relation between the standard three-dimensional human skeleton nodes imported in the motion sensor that connects and the step 3);
5) with step 2) in the motion sensor that connects be placed into the human body corresponding site, by posture alignment motion sensor and standard three-dimensional human skeleton are bound;
6) drive standard three-dimensional human skeleton towards data with the kinematic method of forward direction with what the good motion sensor of binding obtained, export rough demonstration performance campaign;
7) method that the demonstration performance campaign of output is divided with the subspace is carried out feature extraction and segmentation, removes the noise in the demonstration performance motion, extracts motion feature wherein, the output characteristic symbol sebolic addressing;
8) being written into the motion index, is input with the characteristic symbol sequence, carries out retrieval by header, gets rid of dissimilar motion, the output candidate set of moving;
9) set the weight of characteristic node and the number of quantitative result for retrieval, the motion feature sequence vector of importing original demonstration campaign or being converted to by original demonstration motion characteristics symbol sebolic addressing is quantitatively retrieved on the candidate moves collection, obtains net result.
Described method of dividing with the subspace makes up the motion index for the high capacity motion database, each exercise data in the motion database is carried out feature extraction and segmentation, and be converted into subspace characteristic symbol sequence step: come index building with chest, left elbow, right elbow, left wrist, right wrist, left knee, right knee, left ankle, 9 characteristic nodes of right ankle, the local space of the father node of each characteristic node is divided into upper, middle and lower-ranking in vertical direction; Be divided in the horizontal direction front, rear, left and right, in five parts, mark off 15 sub spaces altogether, can represent with a numeric character when characteristic node is in certain sub spaces; Each frame of exercise data is calculated the symbolic representation of these 9 characteristic nodes, and the identical successive frame of symbolic representation is merged into a motion segmentation; Each motion sequence in the motion database is carried out feature extraction, obtain the symbol sebolic addressing after its segmentation.
Described with step 2) in the motion sensor that connects be placed into the human body corresponding site, by posture alignment motion sensor is bound with standard three-dimensional human skeleton: the user makes the initial posture identical with standard three-dimensional human skeleton, system with motion sensor towards the inverse of a matrix matrix multiply by corresponding skeleton node initially towards matrix, obtain the posture alignment matrix; When obtaining demonstration performance, what at first motion sensor will be obtained multiply by the posture alignment matrix towards matrix, just can assign it to corresponding skeleton node then.
The described motion index that is written into, with the characteristic symbol sequence is input, carries out retrieval by header, gets rid of dissimilar motion, the output candidate set of moving: the characteristic node of selecting according to the user carries out dynamic motion segmentation merging to existing motion index, generates new motion index and is loaded into internal memory; Set up corresponding relation between the motion feature sequence to be compared with the method for time series data coupling, and judge both whether correspondent equals; Matching condition between the characteristic symbol is whether the query characteristics symbol is contained in the set of being made up of candidate feature symbol and adjacent feature symbol thereof.
Target of the present invention is intuitive and accurate expression cartoon making personnel's an initiative intention, realizes effectively extracting fast of required exercise data, thereby reuses existing motion material efficiently.Comprise abundant motion material in the high capacity motion database, the animation teacher can obtain the demonstration performance motion easily by motion sensor, and the motion index by means of we make up can rapidly and efficiently retrieve high-quality exercise data similarly.Present existing motion retrieval method or retrieval time is too very long, the result is not accurate enough, otherwise retrieval flow is too loaded down with trivial details, directly perceived inadequately alternately, the present invention has overcome this two difficult problems.In animation process, the 3 d human motion of animation Shi Suoxu generally is the motion capture device acquisition by costliness, and this process need is by the action of professional actor animation teacher design.The present invention allows the animation teacher fully to reuse existing motion material, and does not need each action all to catch again, has saved cartoon making time and cost greatly.The invention solves the problem that the high capacity motion database is difficult to retrieve, the intuitive and efficient exercise data search method of a kind of animation teacher easy to understand and use is provided.
Description of drawings
Fig. 1 (a) is the standard three-dimensional human skeleton signal;
Fig. 1 (b) is the rest signal of motion sensor on human body
Fig. 2 is tied to the standard three-dimensional human skeleton signal by posture alignment with motion sensor;
Fig. 3 is this motion retrieval system framework and flow process signal;
Fig. 4 (a) is the top view signal that the subspace is divided;
Fig. 4 (b) is the side view signal that the subspace is divided;
Fig. 5 is based on the motion feature of dividing the subspace and extracts and the segmentation signal;
Fig. 6 is motion index size and the signal of characteristic node number relation;
Fig. 7 is the signal of motion retrieval example process.
Embodiment
Intuitive and efficient three-dimensional human body movement data retrieval method based on demonstration performance comprises the steps, sees Fig. 3:
1) method of dividing with the subspace, see Fig. 4 (a)-4 (b), for the high capacity motion database makes up the motion index, each exercise data in the motion database is carried out feature extraction and segmentation, and be converted into subspace characteristic symbol sequence, the size of index is relevant with the number of the feature articulation point of selecting, the joint is counted many more, and segmentation is thin more, and index takes up space also just big more, Fig. 6 has showed the relation of motion index size with the characteristic node number, wherein upper and lower F
1, F
2Represent respectively two different nodes are extracted the fragment sequence that feature obtains, middle F extracts the fragment sequence that feature obtains to these two nodes simultaneously, and as seen, the characteristic node number of selecting for use is many more, and motion segmentation is thin more, and required storage space is also big more;
2) connect motion sensor, set the overall reference frame of motion sensor;
3) standard three-dimensional human skeleton of input consumer premise justice is seen Fig. 1 (a);
4) mutual or come given step 2 by program interface by the input configuration file) in corresponding relation between the standard three-dimensional human skeleton nodes imported in the motion sensor that connects and the step 3);
5) with step 2) in the motion sensor that connects be placed into the human body corresponding site, see Fig. 1 (b), by posture alignment motion sensor and standard three-dimensional human skeleton are bound, see Fig. 2;
6) drive standard three-dimensional human skeleton towards data with the kinematic method of forward direction with what the good motion sensor of binding obtained, export rough demonstration performance campaign;
7) method that the demonstration performance campaign of output is divided with the subspace is carried out feature extraction and segmentation, sees Fig. 5, removes the noise in the demonstration performance motion, extracts motion feature wherein, the output characteristic symbol sebolic addressing;
8) being written into the motion index, is input with the characteristic symbol sequence, carries out retrieval by header, gets rid of dissimilar motion, the output candidate set of moving;
9) weight of setting characteristic node and quantitatively result for retrieval number of parameters, the motion feature sequence vector of importing original demonstration campaign or being converted to by the motion feature symbol sebolic addressing carries out quantitative retrieval on the candidate moves collection, obtain net result.
Described method of dividing with the subspace makes up the motion index for the high capacity motion database, each exercise data in the motion database is carried out feature extraction and segmentation, and be converted into subspace characteristic symbol sequence step: come index building with chest, left elbow, right elbow, left wrist, right wrist, left knee, right knee, left ankle, 9 characteristic nodes of right ankle, the local space of the father node of each characteristic node is divided into upper, middle and lower-ranking in vertical direction; Be divided in the horizontal direction front, rear, left and right, in five parts, mark off 15 sub spaces altogether, can represent with a numeric character when characteristic node is in certain sub spaces; Each frame of exercise data is calculated the symbolic representation of these 9 characteristic nodes, and the identical successive frame of symbolic representation is merged into a motion segmentation; Each motion sequence in the motion database is carried out feature extraction, obtain the symbol sebolic addressing after its segmentation.
Described with step 2) in the motion sensor that connects be placed into the human body corresponding site, by posture alignment motion sensor is bound with standard three-dimensional human skeleton: the user makes the initial posture identical with standard three-dimensional human skeleton, system with motion sensor towards the inverse of a matrix matrix multiply by corresponding skeleton node initially towards matrix, obtain the posture alignment matrix; When obtaining demonstration performance, what at first motion sensor will be obtained multiply by the posture alignment matrix towards matrix, just can assign it to corresponding skeleton node then.
The described motion index that is written into, with the characteristic symbol sequence is input, carries out retrieval by header, gets rid of dissimilar motion, the output candidate set of moving: the characteristic node of selecting according to the user carries out dynamic motion segmentation merging to existing motion index, generates new motion index and is loaded into internal memory; Set up corresponding relation between the motion feature sequence to be compared with the method for time series data coupling, and judge both whether correspondent equals; Matching condition between the characteristic symbol is whether the query characteristics symbol is contained in the set of being made up of candidate feature symbol and adjacent feature symbol thereof.
With an example embodiment is described below:
At first the high capacity motion database is carried out pre-service, each exercise data is carried out feature extraction and segmentation, make up the motion index, can carry out seeing Fig. 3 according to flow process then based on the intuitive and efficient three-dimensional human body movement data retrieval of putting on a demonstration of.Connect motion sensor, set its overall reference frame, the input standard three-dimensional human skeleton, see Fig. 1 (a), if the user only pays close attention to local body motion, as motion above the waist, then as long as selected chest, left side elbow, right elbow, left side wrist, 5 characteristic nodes of right wrist, utilize 5 motion sensors can obtain demonstration performance campaign above the waist, come the corresponding relation of designated movement sensor and articulation point alternately or by the input configuration file, then motion sensor is positioned over the human body corresponding site by program interface, see Fig. 1 (b), through after the posture alignment, see Fig. 2, motion sensor promptly can the Real Time Drive human skeleton motion.With the method that the subspace is divided, see Fig. 4 (a)-4 (b), feature extraction is carried out in the demonstration performance motion, obtain its characteristic symbol sequence.The motion index is being loaded in the process of internal memory, the characteristic node system that selectes according to the user carries out dynamic motion segmentation merging to existing motion index, generate new motion index, Fig. 6 has showed this process: suppose that F extracts the motion index sequence that obtains after the feature to certain two node simultaneously, if the user only selects node 1 to carry out motion retrieval, then system's index sequence of carrying out automatically obtaining after segmentation merges is F
1By Fig. 5 we as can be seen, after feature extraction and segmentation, exercise data in demonstration performance motion and the motion database similarly can be exchanged into identical characteristic symbol sequence, therefore can get rid of and the dissimilar exercise data of demonstration performance motion based on the retrieval by header of characteristic sequence, obtain the set of candidate's motion segment, but because the ambiguity of retrieval by header, the motion number of slices that retrieval is come out is many, though comprised all motions similar to input motion, but also comprised some and the dissimilar motion of input motion, therefore needed further accurately retrieval.Set the weight of each feature articulation point and quantitatively after the number of result for retrieval, carry out quantitative retrieval based on dynamic time adjustment (Dynamic Time Warping) algorithm.Finally, we have obtained needed high-quality 3 d human motion data, and the flow process of this retrieval example is seen Fig. 7.
Claims (4)
1. the intuitive and efficient three-dimensional human body movement data retrieval method based on demonstration performance is characterized in that comprising the steps:
1) method of dividing with the subspace makes up the motion index for the high capacity motion database, each exercise data in the motion database is carried out feature extraction and segmentation, and be converted into subspace characteristic symbol sequence;
2) connect motion sensor, set the overall reference frame of motion sensor;
3) standard three-dimensional human skeleton of input consumer premise justice;
4) mutual or come given step 2 by program interface by the input configuration file) in corresponding relation between the standard three-dimensional human skeleton nodes imported in the motion sensor that connects and the step 3);
5) with step 2) in the motion sensor that connects be placed into the human body corresponding site, by posture alignment motion sensor and standard three-dimensional human skeleton are bound;
6) drive standard three-dimensional human skeleton towards data with the kinematic method of forward direction with what the good motion sensor of binding obtained, export rough demonstration performance campaign;
7) method that the demonstration performance campaign of output is divided with the subspace is carried out feature extraction and segmentation, removes the noise in the demonstration performance motion, extracts motion feature wherein, the output characteristic symbol sebolic addressing;
8) being written into the motion index, is input with the characteristic symbol sequence, carries out retrieval by header, gets rid of dissimilar motion, the output candidate set of moving;
9) set the weight of characteristic node and the number of quantitative result for retrieval, the motion feature sequence vector of importing original demonstration campaign or being converted to by original demonstration motion characteristics symbol sebolic addressing is quantitatively retrieved on the candidate moves collection, obtains net result.
2. a kind of intuitive and efficient three-dimensional human body movement data retrieval method as claimed in claim 1 based on demonstration performance, it is characterized in that, described method of dividing with the subspace makes up the motion index for the high capacity motion database, each exercise data in the motion database is carried out feature extraction and segmentation, and be converted into subspace characteristic symbol sequence step: use chest, left side elbow, right elbow, left side wrist, right wrist, left side knee, right knee, left side ankle, 9 characteristic nodes of right ankle come index building, the local space of the father node of each characteristic node is divided in vertical direction, in, following three layers, before being divided in the horizontal direction, after, a left side, right, in five parts, mark off 15 sub spaces altogether, can represent with a numeric character when characteristic node is in certain sub spaces; Each frame of exercise data is calculated the symbolic representation of these 9 characteristic nodes, and the identical successive frame of symbolic representation is merged into a motion segmentation; Each motion sequence in the motion database is carried out feature extraction, obtain the symbol sebolic addressing after its segmentation.
3. a kind of intuitive and efficient three-dimensional human body movement data retrieval method as claimed in claim 1 based on demonstration performance, it is characterized in that, described with step 2) in the motion sensor that connects be placed into the human body corresponding site, by posture alignment motion sensor is bound with standard three-dimensional human skeleton: the user makes the initial posture identical with standard three-dimensional human skeleton, system with motion sensor towards the inverse of a matrix matrix multiply by corresponding skeleton node initially towards matrix, obtain the posture alignment matrix; When obtaining demonstration performance, what at first motion sensor will be obtained multiply by the posture alignment matrix towards matrix, just can assign it to corresponding skeleton node then.
4. a kind of intuitive and efficient three-dimensional human body movement data retrieval method as claimed in claim 1 based on demonstration performance, it is characterized in that, the described motion index that is written into, with the characteristic symbol sequence is input, carry out retrieval by header, get rid of dissimilar motion, the output candidate set of moving: the characteristic node of selecting according to the user carries out dynamic motion segmentation merging to existing motion index, generates new motion index and is loaded into internal memory; Set up corresponding relation between the motion feature sequence to be compared with the method for time series data coupling, and judge both whether correspondent equals; Matching condition between the characteristic symbol is whether the query characteristics symbol is contained in the set of being made up of candidate feature symbol and adjacent feature symbol thereof.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN200810061672A CN100589105C (en) | 2008-05-23 | 2008-05-23 | Visual and efficient three-dimensional human body movement data retrieval method based on demonstrated performance |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN200810061672A CN100589105C (en) | 2008-05-23 | 2008-05-23 | Visual and efficient three-dimensional human body movement data retrieval method based on demonstrated performance |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101308500A true CN101308500A (en) | 2008-11-19 |
| CN100589105C CN100589105C (en) | 2010-02-10 |
Family
ID=40124958
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN200810061672A Expired - Fee Related CN100589105C (en) | 2008-05-23 | 2008-05-23 | Visual and efficient three-dimensional human body movement data retrieval method based on demonstrated performance |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN100589105C (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102129551A (en) * | 2010-02-16 | 2011-07-20 | 微软公司 | Gesture detection based on joint skipping |
| CN103336953A (en) * | 2013-07-05 | 2013-10-02 | 深圳市中视典数字科技有限公司 | Movement judgment method based on body sensing equipment |
| CN103678859A (en) * | 2012-09-12 | 2014-03-26 | 财团法人工业技术研究院 | Motion comparison method and motion comparison system |
| JP2015226757A (en) * | 2014-04-10 | 2015-12-17 | ザ・ボーイング・カンパニーＴｈｅＢｏｅｉｎｇ Ｃｏｍｐａｎｙ | Identifying movements using motion sensing device coupled with associative memory |
| CN106960032A (en) * | 2017-03-21 | 2017-07-18 | 中国科学院深圳先进技术研究院 | 3D shape expression and device |
| CN112349190A (en) * | 2019-08-09 | 2021-02-09 | 天津天堰科技股份有限公司 | Massage training device |
| CN116993872A (en) * | 2023-07-27 | 2023-11-03 | 湖北工程学院 | Labanotation-based human body animation generation system, method, equipment and storage medium |
-
2008
- 2008-05-23 CN CN200810061672A patent/CN100589105C/en not_active Expired - Fee Related
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102129551A (en) * | 2010-02-16 | 2011-07-20 | 微软公司 | Gesture detection based on joint skipping |
| US8633890B2 (en) | 2010-02-16 | 2014-01-21 | Microsoft Corporation | Gesture detection based on joint skipping |
| CN103678859A (en) * | 2012-09-12 | 2014-03-26 | 财团法人工业技术研究院 | Motion comparison method and motion comparison system |
| CN103336953A (en) * | 2013-07-05 | 2013-10-02 | 深圳市中视典数字科技有限公司 | Movement judgment method based on body sensing equipment |
| CN103336953B (en) * | 2013-07-05 | 2016-06-01 | 深圳市中视典数字科技有限公司 | A kind of method passed judgment on based on body sense equipment action |
| JP2015226757A (en) * | 2014-04-10 | 2015-12-17 | ザ・ボーイング・カンパニーＴｈｅＢｏｅｉｎｇ Ｃｏｍｐａｎｙ | Identifying movements using motion sensing device coupled with associative memory |
| CN106960032A (en) * | 2017-03-21 | 2017-07-18 | 中国科学院深圳先进技术研究院 | 3D shape expression and device |
| CN106960032B (en) * | 2017-03-21 | 2021-02-19 | 中国科学院深圳先进技术研究院 | Three-dimensional shape expression method and device |
| CN112349190A (en) * | 2019-08-09 | 2021-02-09 | 天津天堰科技股份有限公司 | Massage training device |
| CN116993872A (en) * | 2023-07-27 | 2023-11-03 | 湖北工程学院 | Labanotation-based human body animation generation system, method, equipment and storage medium |
| CN116993872B (en) * | 2023-07-27 | 2024-01-30 | 湖北工程学院 | Labanotation-based human body animation generation system, method, equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN100589105C (en) | 2010-02-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN100589105C (en) | Visual and efficient three-dimensional human body movement data retrieval method based on demonstrated performance | |
| US10176404B2 (en) | Recognition of a 3D modeled object from a 2D image | |
| CN101138233B (en) | Method for selecting parts of an audiovisual program and device therefor | |
| Torralba et al. | Labelme: Online image annotation and applications | |
| Krüger et al. | Fast local and global similarity searches in large motion capture databases. | |
| CN104850633B (en) | A kind of three-dimensional model searching system and method based on the segmentation of cartographical sketching component | |
| US7356777B2 (en) | System and method for providing a dynamic user interface for a dense three-dimensional scene | |
| Wang et al. | 3D human motion editing and synthesis: A survey | |
| CN105354593B (en) | A kind of threedimensional model sorting technique based on NMF | |
| Afzal et al. | Visualization and visual analytics approaches for image and video datasets: A survey | |
| CN105868706A (en) | Method for identifying 3D model based on sparse coding | |
| CN107958067A (en) | It is a kind of based on without mark Automatic Feature Extraction extensive electric business picture retrieval system | |
| Zhao et al. | Character‐object interaction retrieval using the interaction bisector surface | |
| CN113204636B (en) | Knowledge graph-based user dynamic personalized image drawing method | |
| CN109308324A (en) | A kind of image search method and system based on hand drawing style recommendation | |
| CN108572992A (en) | A kind of method and device of commodity sequence | |
| Yuan et al. | Shrec’19 track: Extended 2D scene sketch-based 3D scene retrieval | |
| Fang et al. | Narrative collage of image collections by scene graph recombination | |
| Wang et al. | iMap: a stable layout for navigating large image collections with embedded search | |
| Baak et al. | An efficient algorithm for keyframe-based motion retrieval in the presence of temporal deformations | |
| Voulodimos et al. | Kinematics-based extraction of salient 3D human motion data for summarization of choreographic sequences | |
| Priya et al. | A comprehensive review of significant researches on content based indexing and retrieval of visual information | |
| CN107066547A (en) | A kind of method for searching three-dimension model based on LIRE | |
| CN110516112A (en) | A kind of human action search method and equipment based on hierarchical model | |
| CN109857886A (en) | A kind of method for searching three-dimension model approached based on minimax value theory of games view |
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 | ||
| C17 | Cessation of patent right | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20100210 Termination date: 20130523 |