CN101441776B - Three-dimensional human body motion editing method driven by demonstration show based on speedup sensor - Google Patents
Three-dimensional human body motion editing method driven by demonstration show based on speedup sensor Download PDFInfo
- Publication number
- CN101441776B CN101441776B CN2008101625955A CN200810162595A CN101441776B CN 101441776 B CN101441776 B CN 101441776B CN 2008101625955 A CN2008101625955 A CN 2008101625955A CN 200810162595 A CN200810162595 A CN 200810162595A CN 101441776 B CN101441776 B CN 101441776B
- Authority
- CN
- China
- Prior art keywords
- action
- training sample
- sample
- acceleration
- demonstration performance
- 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.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 39
- 230000033001 locomotion Effects 0.000 title claims abstract description 35
- 230000001133 acceleration Effects 0.000 claims abstract description 53
- 238000012549 training Methods 0.000 claims abstract description 51
- 230000009471 action Effects 0.000 claims description 85
- 238000010606 normalization Methods 0.000 claims description 6
- 230000011218 segmentation Effects 0.000 claims description 6
- 238000000605 extraction Methods 0.000 claims description 5
- 238000012952 Resampling Methods 0.000 claims description 3
- 238000002474 experimental method Methods 0.000 claims description 3
- 230000005484 gravity Effects 0.000 claims description 3
- 238000007781 pre-processing Methods 0.000 abstract 2
- 239000012634 fragment Substances 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000004080 punching Methods 0.000 description 3
- 238000005070 sampling Methods 0.000 description 3
- 230000010354 integration Effects 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 230000009257 reactivity Effects 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 239000012467 final product Substances 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
Images
Landscapes
- Processing Or Creating Images (AREA)
Abstract
The invention discloses a method for arranging three-dimensional human body movements, which is driven by a demonstration performance and is based on an acceleration sensor. The method comprises the following steps: 1) segmenting movement capturing data and calculating the data segments to obtain acceleration data of an appointed part; 2) generating a training sample set automatically through adding noises; 3) preprocessing the training sample; 4) training a hidden Markov model; 5) connecting the acceleration sensor and placing the acceleration sensor at an appointed part on a human body; 6) performing demonstration movements by a user; 7) preprocessing a demonstration performance sample; 8) identifying the demonstration performance movements by using the hidden Markov model; and 9) editing and modifying the identified movements to obtain a result movement sequence. The method achieves the function of performing movement identification and movement arrangement by using the acceleration sensor, solves the problem that the prior movement identifying method needs to collect a large amount of training samples, and provides a method driven by demonstration performance to perform the movement identification and the movement arrangement on an acceleration space.
Description
Technical field
The present invention relates to 3 D human body action recognition and choreography method, especially a kind of 3 D human body choreography method of putting on a demonstration of driving based on acceleration sensor.
Background technology
The choreography that demonstration performance drives is to grow up on the basis of the existing work in a plurality of fields, and these work comprise cartoon making, the minimum gauge point of use or the motion capture of sensor and the action recognition that the user has nothing to do that doll drives.
The cartoon making that doll drives is the animation method that a kind of effective demonstration performance drives, and the action with the performing artist that it can be real-time is converted to computer virtual role's action.This method provides a kind of on-the-spot demonstration and visualization tool for traditional Film Animation, and its technical essential is: under the target roles situation different with ratio with performing artist's size, how effectively with performing artist's action mapping to target roles.But this method needs special input equipment, and this natural control mode can be directly delivered to virtual role with performing artist's individual character, therefore the control computer doll apparatus requires long practice and professional knowledge usually, has only the expert of minority effectively to grasp.
Motion capture to true man is a kind of animation method of demonstration performance driving of alternative doll, but motion capture equipment is very heavy, costs an arm and a leg, and use is very loaded down with trivial details.Therefore many researchers begin to explore how only to carry out quick and effective action estimation approach with a small amount of gauge point or sensor, can be divided into according to the difference of input equipment: based on method and sensor-based method of vision.The general technology route of these class methods is: utilize a small amount of camera, gauge point or sensor to obtain inadequate movable information, retrieval or action recognition are made in enterprising action in jumbo motion capture data storehouse, obtain high-quality action segment, connect and the action hybrid algorithm by means of action, obtain final continuous human action data.
Have multiple machine learning method to can be used for action recognition, Hidden Markov Model (HMM) is a kind of mode identification method of being used widely, and is particularly suitable for the time series with change in time and space is carried out modeling.From practical angle, a shortcoming of conventional machines learning method is: the great amount of samples that needs to collect a plurality of different users is used for training.Therefore, in fields such as man-machine interactions, the recognizer that the user has nothing to do has caused extensive studies interest, and its gordian technique is: how to generate the enough wide training sample of coverage automatically by computing machine.
Motion sensor is a kind of novel device that can catch many-sided kinetic characteristic, and 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.Acceleration transducer is a kind of motion sensor of measuring acceleration, and it is cheap, is very suitable for the cartoon making that daily interactive application and demonstration performance drive.Based on acceleration sensor, we have proposed a kind of 3 D human body choreography method of putting on a demonstration of driving.
Summary of the invention
The purpose of this invention is to provide a kind of 3 D human body choreography method of putting on a demonstration of driving based on acceleration sensor.
Comprise the steps:
1) to the motion capture data segmentation, the segment data that is only contained single action, read the positional information of skeleton node the motion capture data after segmentation, obtain and quicken the positional information of sensing at the human body institute corresponding virtual position of rest, ask second derivative to obtain acceleration value in its world coordinates to it, this numerical value is coupled with the overall situation rotation of multiply by corresponding node after the acceleration of gravity again on vertical component contrary rotation obtains the acceleration value in the sensor local coordinate;
2) be original sample with above-mentioned acceleration information, generate a series of training samples by adding noise;
3) above-mentioned training sample is carried out pre-service, obtain standardized training sample;
4) read in the initial parameter of Hidden Markov Model (HMM), with standardized training sample training Hidden Markov Model (HMM);
5) connect acceleration sensor and it is placed into the appointed part of user's body;
6) user adorns oneself with acceleration sensor and makes the demonstration performance action, controls the starting and ending of demonstration performance action by a button;
7) above-mentioned demonstration performance sample action is carried out pre-service, obtain standardized demonstration performance sample;
8) with the Hidden Markov Model (HMM) identification step 7 that trains in the step 4)) output the demonstration performance sample;
9) edit-modify is carried out in the action that identifies and obtain the result action sequence, the connection mixing that the content of edit-modify comprises the exaggeration editor that the time of action is adjusted, moves and moves.
Described is original sample with above-mentioned acceleration information, generate a series of training sample steps by adding noise: adopt two kinds of noise generation methods of even distribution noise and Gaussian distribution noise, controlled variable is the ratio of the variance of the variance of signal and noise, adopt the ratio of the variance of the variance of different signals and noise to generate training sample and carry out a series of quantitative experiments, find that ratio value when the variance of the variance of signal and noise is 4 between 5 the time, the action recognition rate is the highest.
Described above-mentioned training sample is carried out pre-service, obtain standardized training sample step: adopt the method for " spline interpolation-resampling " that training sample length is carried out normalization, adopt stochastic variable standardization formula that sample magnitude is carried out normalization, adopt linear pivot analysis algorithm that training sample is carried out feature extraction.
Describedly edit-modify is carried out in the action that identifies obtain the result action sequence, the content of edit-modify comprises exaggeration editor that the time of action is adjusted, moves and the connection blend step that moves: on the acceleration space, utilize the dynamic time adjustment algorithm to find the optimum matching path of demonstration performance sample and the sample action that identifies, and then by increasing or deleting the time attribute that some action action frame is adjusted the action that is identified; From demonstration performance action, extract the dynamics parameter, ask first order derivative to obtain its velocity information, multiply by velocity information after velocity information obtains exaggerating, the action after can obtaining exaggerating to its integration again with the dynamics parameter to the positional information of appointed part; The action that will identify at last is connected mixing with the last action that identifies, and generates result action sequence and output.
Target of the present invention provides a kind of method of directly perceived, natural expression cartoon making personnel creation intention, realizes the layout of 3 D human body action data efficiently.Animation method apparatus expensive at traditional demonstration performance driving, the problem that is difficult to use, we utilize cheap acceleration transducer, proposed a kind of simple and practical choreography method, the user only need degree of will speed up sensor be placed into the assigned address of health and performs to move and gets final product.Need the user to gather the shortcoming of a large amount of training samples at traditional machine learning method, our method can generate all training samples automatically from motion capture data, the action recognition ability of system can be expanded by add new motion capture data in action database.After the demonstration performance that identifies the user, the motion capture data that time attribute during system can also move according to demonstration performance and dynamics attribute obtain identification is made further adjustment.The inventive method does not have particular restriction for the user, even the beginner who lacks experience also can be familiar with and suitable native system in a short period of time, the invention solves the problem that animation method that demonstration performance drives has only the veteran expert of minority just can use, effectively saved manpower and financial resources for the cartoon making industry.
Description of drawings
The configuration signal of Fig. 1 acceleration sensor on human body;
This choreography of Fig. 2 system framework and flow process signal;
Fig. 3 (a) motion capture data fragment example;
The acceleration samples that Fig. 3 (b) obtains from the motion capture data computation;
Fig. 3 (c) is by adding the sample that equally distributed noise generates;
The sample that Fig. 3 (d) generates by the noise that adds Gaussian distribution;
Fig. 3 (e) puts on a demonstration of action;
The sample that Fig. 3 (f) obtains the demonstration performance sampling;
Fig. 4 (a) is through pretreated training sample;
Fig. 4 (b) is through pretreated demonstration performance sample;
The training of Fig. 5 Hidden Markov Model (HMM) and identification signal;
Fig. 6 dynamic time adjustment signal;
Fig. 7 (a) punching action;
Fig. 7 (b) is by the action of the reactivity after hitting;
Punching action after Fig. 7 (c) exaggeration;
Reactivity action after Fig. 7 (d) exaggeration;
Fig. 8 (a) choreography example above the waist;
The choreography example of Fig. 8 (b) lower part of the body.
Embodiment
The 3 D human body choreography method (see figure 2) based on acceleration sensor that demonstration performance drives comprises the steps:
1) to the motion capture data segmentation, the segment data that is only contained single action, read the positional information of skeleton node the motion capture data after segmentation, obtain and quicken the positional information of sensing at the human body institute corresponding virtual position of rest, ask second derivative to obtain acceleration value in its world coordinates to it, this numerical value is coupled with the overall situation rotation of multiply by corresponding node after the acceleration of gravity again on vertical component contrary rotation obtains the acceleration value in the sensor local coordinate;
2) be original sample with above-mentioned acceleration information, generate a series of training samples by adding noise;
3) above-mentioned training sample is carried out pre-service, obtain standardized training sample;
4) read in the initial parameter of Hidden Markov Model (HMM), with standardized training sample training Hidden Markov Model (HMM);
5) connect acceleration sensor and it is placed into the appointed part (see figure 1) of user's body;
6) user adorns oneself with acceleration sensor and makes the demonstration performance action, controls the starting and ending of demonstration performance action by a button;
7) above-mentioned demonstration performance sample action is carried out pre-service, obtain standardized demonstration performance sample;
8) with the Hidden Markov Model (HMM) identification step 7 that trains in the step 4)) output the demonstration performance sample;
9) edit-modify is carried out in the action that identifies and obtain the result action sequence, the connection mixing that the content of edit-modify comprises the exaggeration editor that the time of action is adjusted, moves and moves.
Described is original sample with above-mentioned acceleration information, generate a series of training sample steps by adding noise: adopt two kinds of noise generation methods of even distribution noise and Gaussian distribution noise, controlled variable is the ratio of the variance of the variance of signal and noise, adopt the ratio of the variance of the variance of different signals and noise to generate training sample and carry out a series of quantitative experiments, find that ratio value when the variance of the variance of signal and noise is 4 between 5 the time, the action recognition rate is the highest.
Described above-mentioned training sample is carried out pre-service, obtain standardized training sample step: when different users does same action, speed differs greatly, the acceleration sampled data length that obtains is uneven, adopts the method for " cubic spline interpolation-resampling " that training sample length is carried out normalization; When different users did same action, dynamics differed greatly, and the acceleration sampled data amplitude difference that obtains is very big, adopted that stochastic variable standardization formula carries out normalization to sample magnitude in the theory of probability; The principal character that sample action implied is submerged in the incoherent noise, adopts linear pivot analysis algorithm that training sample is carried out feature extraction, just is used for training and identification through the training sample after the pre-service.
Describedly edit-modify is carried out in the action that identifies obtain the result action sequence, the content of edit-modify comprises exaggeration editor that the time of action is adjusted, moves and the connection blend step that moves: on the acceleration space, utilize the dynamic time adjustment algorithm to find the optimum matching path of demonstration performance sample and the sample action that identifies, and then by increasing or deleting the time attribute that some action action frame is adjusted the action that is identified; From demonstration performance action, extract the dynamics parameter, ask first order derivative to obtain its velocity information, multiply by velocity information after velocity information obtains exaggerating, the action after can obtaining exaggerating to its integration again with the dynamics parameter to the positional information of appointed part; The action that will identify at last is connected mixing with the last action that identifies, and generates result action sequence and output.
With an example embodiment is described below:
At first each the action fragment in the motion capture data storehouse is calculated the acceleration information of corresponding body part as initial sample, Fig. 3 (a) is a motion capture data fragment example, and Fig. 3 (b) is the acceleration samples that calculates from this motion capture data.Then, system generates a series of training samples automatically by adding noise, has two kinds of noise generation methods available: the noise that evenly distributes, see Fig. 3 (c), and the Gaussian distribution noise is seen Fig. 3 (d).Training sample set is carried out standardization and feature extraction, and Fig. 4 (a) is through a training sample example after the standardization.Next step, with training Hidden Markov Model (HMM) through the sample set after the pre-service, training process is seen Fig. 5.Connect acceleration sensor and it is placed into the appointed part of human body, Fig. 1 has showed the rest of acceleration sensor on human body.Then, the user adorns oneself with acceleration sensor and makes the demonstration performance action, control the starting and ending that demonstration performance is moved by a button, obtain the acceleration sampling, Fig. 3 (e), Fig. 3 (f) are a demonstration performance action and the sample that the acceleration sensor sampling obtains.Before identification demonstration performance action, also need it is carried out standardization and feature extraction, Fig. 4 (b) is through one after standardization demonstration performance action example.With the Hidden Markov Model (HMM) identification demonstration performance action that trains, its process is seen Fig. 5.At last, according to the time attribute of demonstration performance action the dynamic time adjustment is carried out in the action that identifies, Fig. 6 is that dynamic time is adjusted example, and wherein, action data is illustrated with one one dimension curve; Dynamics according to demonstration performance action is exaggerated motion editing to the action that identifies, and Fig. 7 (a)-(d) has showed the effect of exaggeration is moved in punching action and reaction action thereof respectively.The 3 D human body choreography example that Fig. 8 (a) and Fig. 8 (b) have then provided respectively above the waist and lower part of the body demonstration performance drives based on acceleration sensor.
Claims (1)
1. the 3 D human body choreography method based on acceleration sensor of putting on a demonstration of driving is characterized in that comprising the steps:
1) to motion capture data segmentation in the motion capture data storehouse, the segment data that is only contained single action, read the positional information of skeleton node the motion capture data after segmentation, obtain and the positional information of acceleration sensor at the human body institute corresponding virtual position of rest, ask second derivative to obtain acceleration value in its world coordinates to it, acceleration value is coupled with the contrary acceleration value that obtains in the acceleration sensor local coordinate of rotating that multiply by the overall situation rotation of corresponding node after the acceleration of gravity again on vertical component in this world coordinates;
2) be original sample with the acceleration value in the above-mentioned local coordinate, generate a series of training samples by adding noise;
3) above-mentioned training sample is carried out pre-service, obtain standardized training sample;
4) read in the initial parameter of Hidden Markov Model (HMM), with standardized training sample training Hidden Markov Model (HMM);
5) connect acceleration sensor and it is placed into the appointed part of user's body;
6) user adorns oneself with acceleration sensor and makes the demonstration performance action, controls the starting and ending of demonstration performance action by a button;
7) sample to above-mentioned demonstration performance action carries out pre-service, obtains standardized demonstration performance sample;
8) with the Hidden Markov Model (HMM) identification step 7 that trains in the step 4)) output standardized demonstration performance sample;
9) edit-modify is carried out in the action that identifies and obtain the result action sequence, the connection mixing that the content of edit-modify comprises the exaggeration editor that the time of action is adjusted, moves and moves;
Described is original sample with the acceleration value in the above-mentioned local coordinate, generate a series of training sample steps by adding noise: adopt two kinds of noise generation methods of even distribution noise or Gaussian distribution noise, controlled variable is the ratio of the variance of the variance of signal and noise, adopt different described controlled variable to generate training sample and carry out a series of quantitative experiments, when the ratio value of the variance of the variance of signal and noise is 4 between 5 the time, the action recognition rate is the highest;
Described above-mentioned training sample is carried out pre-service, obtain standardized training sample step: adopt the method for " spline interpolation-resampling " that training sample length is carried out normalization, adopt stochastic variable standardization formula that the training sample amplitude is carried out normalization, adopt linear pivot analysis algorithm that training sample is carried out feature extraction.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2008101625955A CN101441776B (en) | 2008-12-04 | 2008-12-04 | Three-dimensional human body motion editing method driven by demonstration show based on speedup sensor |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2008101625955A CN101441776B (en) | 2008-12-04 | 2008-12-04 | Three-dimensional human body motion editing method driven by demonstration show based on speedup sensor |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101441776A CN101441776A (en) | 2009-05-27 |
| CN101441776B true CN101441776B (en) | 2010-12-29 |
Family
ID=40726195
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2008101625955A Expired - Fee Related CN101441776B (en) | 2008-12-04 | 2008-12-04 | Three-dimensional human body motion editing method driven by demonstration show based on speedup sensor |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101441776B (en) |
Families Citing this family (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8947441B2 (en) | 2009-06-05 | 2015-02-03 | Disney Enterprises, Inc. | System and method for database driven action capture |
| CN101615301B (en) * | 2009-07-29 | 2013-03-27 | 腾讯科技（深圳）有限公司 | Path control method and system for target in computer virtual environment |
| CN101989076A (en) * | 2010-08-24 | 2011-03-23 | 北京水晶石数字科技有限公司 | Method for controlling shooting by three-dimensional software |
| CN101995835B (en) * | 2010-08-24 | 2012-06-27 | 北京水晶石数字科技股份有限公司 | System for controlling performance by three-dimensional software |
| CN101989079A (en) * | 2010-08-24 | 2011-03-23 | 北京水晶石数字科技有限公司 | System for controlling photography by three-dimensional software |
| CN101976330B (en) * | 2010-09-26 | 2013-08-07 | 中国科学院深圳先进技术研究院 | Gesture recognition method and system |
| CN101976451B (en) * | 2010-11-03 | 2012-10-03 | 北京航空航天大学 | Motion control and animation generation method based on acceleration transducer |
| CN102722312B (en) * | 2011-12-16 | 2015-12-16 | 江南大学 | A kind of action trend prediction method of interaction experience based on pressure transducer and system |
| CN104463090A (en) * | 2013-11-25 | 2015-03-25 | 安徽寰智信息科技股份有限公司 | Method for recognizing actions of human body skeleton of man-machine interactive system |
| WO2016123648A1 (en) * | 2015-02-02 | 2016-08-11 | Guided Knowledge Ip Pty Ltd | Frameworks, devices and methodologies configured to enable delivery of interactive skills training content, including content with multiple selectable expert knowledge variations |
| US10918924B2 (en) | 2015-02-02 | 2021-02-16 | RLT IP Ltd. | Frameworks, devices and methodologies configured to enable delivery of interactive skills training content, including content with multiple selectable expert knowledge variations |
| JP6836582B2 (en) | 2015-05-08 | 2021-03-03 | ジーエヌ アイピー ピーティーワイ リミテッド | Frameworks, devices and methodologies configured to enable automated classification and / or retrieval of media data based on user performance attributes obtained from the performance sensor unit. |
| CN109074752A (en) | 2015-12-10 | 2018-12-21 | Gn股份有限公司 | It is configured as realizing the frame and method of the real-time adaptive transmission of skill training data based on user's performance is monitored by performance monitoring hardware |
| CN108462707B (en) * | 2018-03-13 | 2020-08-28 | 中山大学 | Mobile application identification method based on deep learning sequence analysis |
| CN109269483B (en) * | 2018-09-20 | 2020-12-15 | 国家体育总局体育科学研究所 | Calibration method, calibration system and calibration base station for motion capture node |
| CN110516389B (en) * | 2019-08-29 | 2021-04-13 | 腾讯科技（深圳）有限公司 | Behavior control strategy learning method, device, equipment and storage medium |
-
2008
- 2008-12-04 CN CN2008101625955A patent/CN101441776B/en not_active Expired - Fee Related
Also Published As
| Publication number | Publication date |
|---|---|
| CN101441776A (en) | 2009-05-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101441776B (en) | Three-dimensional human body motion editing method driven by demonstration show based on speedup sensor | |
| CN109685848A (en) | A kind of neural network coordinate transformation method of three-dimensional point cloud and three-dimension sensor | |
| CN101751692A (en) | Method for voice-driven lip animation | |
| CN110065068A (en) | A kind of robotic asssembly operation programming by demonstration method and device based on reverse-engineering | |
| CN103473801A (en) | Facial expression editing method based on single camera and motion capturing data | |
| CN103778661B (en) | A kind of method, system and computer for generating speaker's three-dimensional motion model | |
| CN103823554A (en) | Digital virtual-real interaction system and digital virtual-real interaction method | |
| CN107024989A (en) | A kind of husky method for making picture based on Leap Motion gesture identifications | |
| Nandy et al. | Recognizing & interpreting Indian sign language gesture for human robot interaction | |
| CN105243375A (en) | Motion characteristics extraction method and device | |
| CN105500370A (en) | Robot offline teaching programming system and method based on somatosensory technology | |
| CN111079547B (en) | Pedestrian moving direction identification method based on mobile phone inertial sensor | |
| CN117032453A (en) | Virtual reality interaction system for realizing mutual recognition function | |
| CN111159872A (en) | Three-dimensional assembly process teaching method and system based on human-machine engineering simulation analysis | |
| CN109910004A (en) | User interaction approach, control equipment and storage medium | |
| CN112365580A (en) | Virtual operation demonstration system for human-computer skill teaching | |
| CN1466104A (en) | Statistics and rule combination based phonetic driving human face carton method | |
| CN104463968B (en) | The matching of remote sensing image binocular stereo vision and three-dimensional rebuilding method based on power grid GIS three-dimensional platform | |
| CN106022466A (en) | Personalized robot and method for realizing the personalization of the robot | |
| CN110000753A (en) | User interaction approach, control equipment and storage medium | |
| CN109886109B (en) | Behavior identification method based on deep learning | |
| CN109807898A (en) | Motion control method, control equipment and storage medium | |
| CN1952850A (en) | Three-dimensional face cartoon method driven by voice based on dynamic elementary access | |
| Wang et al. | Robot gaining robust pouring skills through fusing vision and audio | |
| CN107692669A (en) | A kind of artificial intelligence baby accompanies and attends to robot and its method of accompanying and attending to |
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: 20101229 Termination date: 20121204 |