CN105809144B - A kind of gesture recognition system and method using movement cutting - Google Patents
A kind of gesture recognition system and method using movement cutting Download PDFInfo
- Publication number
- CN105809144B CN105809144B CN201610177623.5A CN201610177623A CN105809144B CN 105809144 B CN105809144 B CN 105809144B CN 201610177623 A CN201610177623 A CN 201610177623A CN 105809144 B CN105809144 B CN 105809144B
- Authority
- CN
- China
- Prior art keywords
- gesture
- cutting
- data
- module
- movement
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/20—Movements or behaviour, e.g. gesture recognition
- G06V40/28—Recognition of hand or arm movements, e.g. recognition of deaf sign language
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/01—Input arrangements or combined input and output arrangements for interaction between user and computer
- G06F3/011—Arrangements for interaction with the human body, e.g. for user immersion in virtual reality
- G06F3/012—Head tracking input arrangements
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/20—Movements or behaviour, e.g. gesture recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Psychiatry (AREA)
- Social Psychology (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
- User Interface Of Digital Computer (AREA)
Abstract
A kind of gesture recognition system and method using movement cutting is claimed in the present invention, is related to machine vision and field of human-computer interaction, the described method comprises the following steps: detection head movement first, and calculates head pose variation；Then sliced signal is sent according to Attitude estimation information and judges gesture cutting whole story point, capture gesture sequence of frames of video in the time interval that this gesture executes if signal indicates the cutting of active gesture motion, and pretreatment and feature extraction are carried out to gesture frame image；Sequence of frames of video is acquired in real time if signal indicates automatic gesture motion cutting and automatically analyzes cut-off by analyzing the motion change rule between adjacent gesture, carry out movement cutting, vision extraneous features are extracted by effective first gesture sequence that cutting obtains again, and by having the Gesture Recognition Algorithm for eliminating space and time difference to obtain types results.Present invention substantially reduces the computing costs of the redundancy of continuous gesture and recognizer, improve the accuracy and real-time of gesture identification.
Description
Technical field
The invention belongs to digital pictures and field of human-computer interaction, more particularly, to a kind of gesture identification using movement cutting
System and method.
Background technique
With the development that mobile phone touch operation and human body tracking identify, people, which have realized gesture interaction mode, to be had with people
Centered on naturality, the advantages such as terseness and substantivity, the interactive interface based on manpower intelligent input becoming new skill
Art trend, especially with the rise of immersive VR new equipment, various interaction schemes are used to improve immersion experience,
It is wherein the most succinct with gesture interaction, direct, natural.
Gesture identification is widely used to augmented reality, virtual reality, somatic sensation television game etc. as a kind of human-computer interaction means
Scene, for these application scenarios, operating gesture is randomly-embedded in continuous action stream, current many view-based access control models
Gesture recognition system assumes that between each movement of input there is the independent gesture paused or segmented, and in real-time scene
Under application study it is relatively fewer.
The Chinese invention patent of Publication No. CN102789568B discloses a kind of gesture identification side based on depth information
Method obtains the motion profile of hand tracking, recycles by detecting hand outline information on each independent human region
Hidden Markov Model carries out modeling to motion profile and identifies gesture, in analysis time limit fixed track length in real time to filter hand
Gesture track.Method also is to allow user when executing gesture by clicking button come the starting point and end of cutting gesture
Point.
Hand and non-rigid have flexibility, and the hand in movement has the characteristics that shape is changeable, the same movement of the same person
It is inscribed in different viewing angles or observation, can all there be larger difference in form and duration, and different people is more difficult to reappear
The same movement, therefore had very using which kind of gesture feature description to the effect and universality of view-based access control model gesture identification
It is big to influence.Currently, gesture model has the gesture model based on apparent gesture model and based on 3D model, Publication No.
The Chinese invention patent of CN102880865B provides a kind of dynamic gesture identification method based on the colour of skin and morphological feature, root
Threshold value cutting processing is made to image according to distribution characteristics of the skin color of people in YCrCb color space, and is obtained with ellipse fitting
Manpower colour of skin block is obtained, then marks the center of gravity of color lump, and then judge the movement of hand according to gravity motion, reaches identifying purpose.
In conclusion existing gesture identification exchange method has the disadvantage in that (1) is difficult under the conditions of practical application
Positioning has the beginning and end key point of function definition gesture in complicated gesture stream；(2) same gesture is due to executing speed
There is spatio-temporal difference in different different with movement range and inevitable area, and the accuracy rate and robustness to identification will cause very big
It influences.
Summary of the invention
For the deficiency of the above technology, propose a kind of precision that gesture identification can be improved and efficiency using movement
The gesture recognition system and method for cutting.Technical scheme is as follows: a kind of gesture recognition system using movement cutting,
Comprising: data interface module, data interface module includes gesture data stream interface component and head pose data stream interface；With
In the initial data for obtaining simultaneously management header posture and hand gestures；
Head pose estimation module, including feature calculation module and the predefined module of posture, for head attitude data
The initial data that stream interface is sent handles original header attitude data with algorithm for estimating and output estimation result gives movement segmentation
Module；
Movement segmentation module, comprising automatic segmentation module, actively divides module and buffer area, according to head pose estimation mould
The output estimation of block is as a result, carry out cutting processing to gesture data using segmentation algorithm；
Gesture recognition module, the gesture sequence for obtaining to movement segmentation module cutting carry out gesture identification.
Further, the data interface module is responsible for management header attitude data stream and gesture video requency frame data stream, hand
Gesture acquisition equipment includes that perhaps data glove head pose acquisition equipment includes depth camera or wears biography depth camera
Sensor, the head pose acquisition equipment have programmable application programming interfaces and are able to carry out according to collecting work.
Further, feature used by the head pose estimation module is described as head pose angle, and head pose is adopted
Collection equipment acquisition data through head towards vector calculate after be used to express head pose Euler's rotation angle or face orientation to
Amount, head or face orientation ο (x, y, z) are found out by following formula: ο=z (α) x (β) z (γ) P, wherein z (α) is indicated
Around reference frame z-axis rotation alpha angle value, z (γ) indicates that head rotates γ angle value, x (β) around reference frame z-axis on head
Head is indicated around reference frame x-axis rotation β angle value, P (1,0,0) indicates the point under reference frame.
Further, by the way of being determined based on head pose, head pose refers to fixed in advance the module of actively dividing
The face of justice towards or headwork, to represent head pose that gesture starts as the starting point of cutting, to represent gesture
End point of the head pose of end as cutting；The automatic segmentation module uses the template matching side based on empty-handed potential model
Method judges the inflection point acted between different gestures according to the rate variation of gesture motion variation；The relevant parameter of automatic segmentation has
SPOTTING_START cutting starting point threshold value, SPOTTING_END cutting end point threshold value, DURATION gesture length threshold,
Cutting starting point is marked as rate α (t) the > SPOTTING_START of the characterization hand exercise of a certain moment t, as α (t) <
When SPOTTING_END, cutting end point is marked, the gesture sequence length L (g) that cutting is obtained is compared with DURATION,
If the gesture sequence duration of cutting is too short, the sequence is abandoned.
Further, the gesture recognition module includes characteristic extracting module, disaggregated model module and match cognization mould
Block, characteristic extracting module are used to carry out feature extraction and calculating to the data that movement segmentation obtains, and disaggregated model module is for building
Vertical gesture classification model, the model are trained study to the feature being calculated, and match cognization module is used for defeated according to gesture
Enter the feature learnt in feature and disaggregated model and carry out match cognization, the quantified output of the recognition result is mapped as operation and refers to
It enables.
Further, the match cognization module of the gesture recognition module uses Discrete HMM Hidden Markov Model
Or DTW dynamic time warping sorting algorithm is handled and exports recognition result.
Further, the characteristic extracting module is used using the local quaternary number in gesture joint as feature vector, is calculated
Steps are as follows: first finding out the bone direction vector under local coordinate system, renormalization；And then find out direction vector and three axis units
The angle of vector；After obtaining the angle of three axis, the local quaternary number expression of gesture joint point feature is calculated.
A kind of gesture identification method using movement cutting based on the system comprising following steps:
1) a kind of acquisition equipment of data, is selected first, and establishes data-interface to provide continuous gesture number in real time
According to source；2), detection head athletic posture is first gesture, and is matched with first gesture type predetermined；If 3), detect master
Movement cutting head gesture is then according to cutting head gesture signal calling data-interface acquisition gesture data is started, according to end cutting head gesture
Signal-off data-interface stops data collection；4), data-interface is called to start if detecting auto-action cutting head gesture
Data are acquired, while gesture motion cutting is carried out according to hand exercise state or empty-handed potential model；5), finally with recognizer pair
The gesture sequence that cutting obtains carries out gesture identification.
Further, the step of step 4) automatic segmentation includes:
Continuous data and abstraction templates are subjected to preliminary matches, enable GtIndicate the input gesture of a length of t at one section, essence
It is an eigenmatrix, line number is the hits in time t, and every behavior gesture feature vector, the vector is by the quaternary number spy
Sign is calculated, template library (g1,g2,…,gn) indicate that known n kind is abstracted gesture template, S (Gt,gi) measurement current data stream
With the similitude of certain gesture template；When hand exercise stops, by gesture sequence GtWith template (g1,g2,…,gn) feature into
Row compares, and seeks locally optimal solution by Hill Climbing algorithm, if in GtIn detect known gesture template piece
It, then be compared by section with the template gesture, if detecting similitude, the similitude of the two just be will increase, until occurring
Similitude drop point, that is, inflection point is detected by the end point of movement segment, puts it into alternative collection, to its with similitude
Its gesture template carries out same detection step, then takes whole story point of the maximum segment of similarity measure as cutting in alternative collection, i.e.,
Set metasequence can be converted by input gesture.
It advantages of the present invention and has the beneficial effect that:
The present invention uses that user is unrelated and feature description of the vision extraneous features as gesture, so that different users can be with
With identical gesture identification model do not have to additionally train new gesture template library, by with the movement based on head pose estimation
Cutting method effectively cutting complexity gesture sequence, reduction redundant data can improve gesture identification precision and efficiency.The present invention mentions
The gesture identification method and system using movement cutting out, can be improved the precision and efficiency of gesture identification, can be effective
It solves under real-time scene, the gesture cutting problems between the spatio-temporal difference problem and gesture start and ending of continuous dynamic gesture.
Detailed description of the invention
Fig. 1 is that the present invention provides preferred embodiment gesture recognition system structural block diagram；
Fig. 2 is head pose angle model and embodiment schematic diagram referenced by the present invention；
Fig. 3 is the corresponding head pose estimation method flow diagram of the present invention；
Fig. 4 is the active cutting method flow chart in present invention movement cutting method.
Fig. 5 is the present invention to the automatic segmentation method flow diagram in movement cutting method.
Fig. 6 is gesture identification method flow chart of the present invention using movement cutting.
Specific embodiment
Below in conjunction with attached drawing, the invention will be further described:
A kind of gesture identification method and system using movement cutting, the system specifically include that as shown in Figure 1
A1~A14:A1 be with data acquisition equipment be adapted data interface module, A2 be head pose estimation module,
A3 is that movement cutting module, A4 are gesture recognition modules, wherein data interface module contains head pose data stream interface group
Part A5 and gesture data stream interface component A6, head pose estimation module contain feature calculation component A7 and the predefined mould of posture
Board group part A8, movement cutting module contain automatic segmentation component A9 and active cutting component A10, A11 use in buffer area therein
First gesture segment is stored, gesture feature extraction assembly A12, gesture-type and operational order are contained in gesture recognition module reflects
The disaggregated model component A14 for penetrating component A13, being obtained by training.
By taking immersive VR scene as an example, head pose angle embodiment schematic diagram of the invention is as shown in Figure 2 b, head
Attitude data stream interface corresponding equipment in portion's is the head-mounted display comprising inertial sensor, and such product has had in the market
It sells, and substantially all includes attitude angle inducing function.Head pose angle model schematic diagram is as shown in Figure 2 a, and head coordinate system is opposite to join
Three Euler's angular dependences for examining coordinate system reflect the posture on head: the angle of head coordinate system and referential horizontal plane is pitching
Angle θ (pitch) bows when on the horizontal plane that the positive axis of head coordinate system was located at the origin of reference frame (new line)
The elevation angle is positive, and is otherwise negative.Head coordinate system xbThe projection of axis in the horizontal plane and reference frame xgAngle between axis is inclined
It navigates angle ψ (yaw), by xgAxis rotates counterclockwise to head xbProjection line when, yaw angle is positive, otherwise be negative.Head coordinate system zb
Axis with pass through head xbAngle between the vertical guide of axis is roll angle φ (roll), and the right wryneck in head is positive, otherwise is negative.
Algorithm of Head Pose Estimation process is as shown in Figure 3.
B1~B3: for wearing sensor described in [0023], B1 is to pass through inertial sensor detection and capture head appearance
State variation, and with the data interface management attitude angle data of building, B2 be to acquired attitude angle data progress feature extraction,
Calculation method is as follows:
ο=z (α) x (β) z (γ) P (1)
Wherein, ο (x, y, z) indicate head or face towards vector, z (α) indicate head around referential z-axis rotation alpha angle,
Z (γ) indicates that head rotates γ angle around reference frame z-axis, and x (β) indicates that head rotates around x axis β angle, uses spin matrix
It indicates:
Wherein, head is around the rotation angle [alpha] of each axis, and beta, gamma can be obtained by the data-interface for wearing inertial sensor, by formula
Sub (2)-(5), which are brought into (1) formula, can be obtained head towards vector ο, be used to the head pose feature indicated.
B6~B5: according to head pose vector set Ο (ν predeterminedi) in template νiWith the feature of extract real-time
Vector ο is matched, according to Distance conformability degree principle judge ο posture type and output quantization as a result, finally according to current appearance
Whether state type decision enters cutting or terminates cutting.
A16~A17: by bis- positioning results of sobel, inputting SVM license plate judgment models, obtain real license plate block, and
Output.
B7: at the time of timer is used to record generation cutting posture, counter is used to calculate the cutting in threshold time period
Frequency can expand sliced signal by this two pieces processing, i.e., changed by head pose or acted to generate
Cutting foundation.
Acting cutting is the necessary hand for applying to gesture identification effectively in real-time scene such as immersive VR
Section, unlike the isolated gesture identification under experiment condition, real-time conditions are increasingly complex, are containing excessive gesture and not operation gesture
Continuous data stream, the present invention devise two kinds of cutting methods from validity and real-time, and active syncopation is based on head appearance
State estimation, the subjective initiative of user can be played as gesture and does not interfere the normal operating of hand, its main feature is that with
Centered on family, cutting accuracy is high；Automatic segmentation is the supplement to active cutting, for handling i.e. first gesture of effective gesture or so
For empty-handed gesture or the apparent simple scenario that pauses.
It is illustrated in figure 4 the implementation flow chart of active cutting method:
C1~C5: when continuous gesture stream initial data introduces system by data-interface, head is determined according to the method
For portion's posture to determine cut-off, the data after cutting are exactly first gesture, i.e., the sequence fragment of effective gesture.It is specifically exactly to work as
Detect cutting initial point signal i.e. head pose when call data-interface currently to start data data introduce pending data
In buffer area, relevant interface is then closed when detecting cutting distal tip signal, stops introducing the data of buffer area, cutting
First gesture data to be processed can be obtained in end, which will be delivered to characteristic extracting module and do processing in next step, and actively cut
Sub-module continues to monitor the sliced signal from head pose estimation module.
It is illustrated in figure 5 the method flow diagram of automatic segmentation:
D1~D7: continuous data and abstraction templates carry out preliminary matches, GtIndicate the dynamic gesture of a length of t at one section, mould
Plate library (g1,g2,…,gn) indicate that known n kind is abstracted gesture template, S (Gt,gi) measure current data stream and certain gesture template
Similitude.Both when hand exercise stops, whole string data is compared with template, once detect similitude, then
Similitude will be stepped up, and be detected by the end point of movement segment when inflection point occurs in similitude decline, then take the overall situation
Whole story point of the maximum string of similarity measure as cutting, can convert gesture to set metasequence.
It is illustrated in figure 6 the gesture identification method flow chart using movement cutting:
For E1~E3 by taking immersive VR scene as an example, used head pose input equipment is with inertia sensing
The head-mounted display that can detecte head pose angle of device, gesture input device are the depth cameras that can carry out bone tracking
Machine passes through head pose data-interface real-time detection head pose and converts sliced signal for estimated result, and gesture data connects
Mouth carries out active cutting to gesture with reference to sliced signal, and when such as executing gesture, user is hoped to arm, generation end rotation variation,
Beginning sliced signal is then generated, when user terminates gesture simultaneously while going back to positive head, then generates end sliced signal, this process pair
E9~E12 is answered to can choose if user is not desired to using active mode into automatic gesture cutting mode E8, the mode is to transport
Dynamic inflection point judges dicing position, with reference to cut-off and data-interface is called to introduce metadata streams to data buffer zone.
E4~E5: the metadata obtained by [0031] is subjected to gesture pretreatment, further according to depth camera such as Kinect
The gesture associated joint point data that tracking bone obtains carries out feature extraction and calculation, and the present embodiment is using local the four of gesture joint
For first number as feature vector, calculation method is as follows:
(1) the bone direction vector under local coordinate system is first found out, then unitization.
(2) angle of direction vector and three axis unit vectors is found out again.
(3) after the angle for obtaining three axis, local quaternary number q is calculated.
E6~E7: in order to solve the problems, such as that accuracy of identification caused by spatio-temporal difference declines, the present embodiment uses dynamic time
Regular DTW algorithm carries out model training and match cognization, can be very good to solve the problems, such as that length of time series is unequal, specifically
Process is to give two time serieses: R=< r1,r2,…,rm> and T=< t1,t2,…,tn>, DTW by calculate R and T it
Between fitst water matching φ=(φR,φT) make the distance of corresponding element after matching only and minimum.For Optimum Matching φ, φR
=(φ1 R,φ2 R,…φK R),(1≤φi R≤ m, 1≤i≤K), and φT=(φ1 T,φ2 T,…φK T),(1≤φi T≤n,1
≤i≤K).DTW distance definition between time series R and T is as follows:Wherein, d
(i, j) indicates in R the distance between j-th of element in i-th of element and T, and the present embodiment takes d (i, j)=(Li+Lj)/2, wherein
LiAnd LjIndicate first gesture length.Classify finally by based on the likeness in form degree of distance measurement to gesture, the present embodiment uses
Arest neighbors NN classifier is classified.
The above embodiment is interpreted as being merely to illustrate the present invention rather than limit the scope of the invention.?
After the content for having read record of the invention, technical staff can be made various changes or modifications the present invention, these equivalent changes
Change and modification equally falls into the scope of the claims in the present invention.
Claims (7)
1. a kind of gesture recognition system using movement cutting characterized by comprising data interface module, data-interface mould
Block includes gesture data stream interface component and head pose data stream interface；For obtaining simultaneously management header posture and hand gestures
Initial data；
Head pose estimation module, including feature calculation module and the predefined module of posture, for being connect to head attitude data stream
The initial data that mouth is sent handles original header attitude data with algorithm for estimating and output estimation result gives movement segmentation module；
Movement segmentation module, comprising automatic segmentation module, actively divides module and buffer area, according to head pose estimation module
Output estimation is as a result, carry out cutting processing to gesture data using segmentation algorithm；The module of actively dividing is used based on head
The mode of pose discrimination, head pose refers to face's direction predetermined or headwork, to represent the head that gesture starts
Starting point of the posture as cutting, to represent head pose that gesture terminates as the end point of cutting；The automatic segmentation mould
Block uses the template matching method based on empty-handed potential model, judges to move between different gestures according to the rate variation of gesture motion variation
The inflection point of work；The relevant parameter of automatic segmentation has SPOTTING_START cutting starting point threshold value, SPOTTING_END cutting knot
Beam spot threshold value, DURATION gesture length threshold, as rate α (t) the > SPOTTING_ of the characterization hand exercise of a certain moment t
Cutting starting point is marked when START, as α (t) < SPOTTING_END, marks cutting end point, the gesture that cutting is obtained
Sequence length L (g) is compared with DURATION, if the gesture sequence duration of cutting is too short, abandons the sequence；
Gesture recognition module, the gesture sequence for obtaining to movement segmentation module cutting carry out gesture identification.
2. the gesture recognition system according to claim 1 using movement cutting, which is characterized in that the data-interface mould
Block is responsible for management header attitude data stream and gesture video requency frame data stream, and it includes depth camera or data that gesture, which acquires equipment,
Gloves, head pose acquisition equipment include depth camera or wear sensor, and the head pose acquisition equipment has can
The application programming interfaces of programming are simultaneously able to carry out according to collecting work.
3. the gesture recognition system according to claim 1 using movement cutting, which is characterized in that the gesture identification mould
Block includes characteristic extracting module, disaggregated model module and match cognization module, and characteristic extracting module is used to obtain movement segmentation
Data carry out feature extraction and calculating, disaggregated model module is for establishing gesture classification model, and the model is to being calculated
Feature is trained study, and match cognization module is used to be carried out according to the feature learnt in gesture input feature and disaggregated model
Match cognization, the quantified output of the recognition result are mapped as operational order.
4. the gesture recognition system according to claim 3 using movement cutting, which is characterized in that the gesture identification mould
The match cognization module of block is used to be carried out using Discrete HMM Hidden Markov Model or DTW dynamic time warping sorting algorithm
It handles and exports recognition result.
5. the gesture recognition system according to claim 3 using movement cutting, which is characterized in that the feature extraction mould
For block using the local quaternary number in gesture joint as feature vector, steps are as follows for calculating: first finding out the bone side under local coordinate system
To vector, renormalization；And then find out the angle of direction vector and three axis unit vectors；After obtaining the angle of three axis, calculate
Local quaternary number to gesture artis feature is expressed.
6. it is a kind of based on system described in claim 1 using movement cutting gesture identification method, which is characterized in that including with
Lower step:
1) a kind of acquisition equipment of data, is selected first, and establishes data-interface to provide continuous gesture data source in real time；
2), detection head athletic posture is first gesture, and is matched with first gesture type predetermined；If 3), detected actively dynamic
Make cutting head gesture then according to cutting head gesture signal calling data-interface acquisition gesture data is started, according to end cutting head gesture signal
Data-interface is closed to stop data collection；4), data-interface is called to start to acquire if detecting auto-action cutting head gesture
Data, while gesture motion cutting is carried out according to hand exercise state or empty-handed potential model；5), finally with recognizer to cutting
Obtained gesture sequence carries out gesture identification.
7. the gesture identification method according to claim 6 using movement cutting, which is characterized in that the step 4) is automatic
The step of cutting includes:
Continuous data and abstraction templates are subjected to preliminary matches, GtIndicate the input gesture data sequence of a length of t at one section,
Matter is an eigenmatrix, and line number is the hits in time t, and every behavior gesture feature vector, the vector is by quaternary number feature
It is calculated, template library (g1,g2,…,gn) indicate that known n kind is abstracted gesture template, S (Gt,gi) measurement current data stream and
The similitude of certain gesture template；When hand exercise stops, by the whole string data G of gesture sequencetWith template (g1,g2,…,gn)
It is compared, locally optimal solution is sought by Hill Climbing algorithm, if in GtIn detect known gesture template piece
It, then be compared by section with the template gesture, if detecting similitude, the similitude of the two just be will increase, until occurring
Similitude drop point is detected by the end point of movement segment, puts it into alternative collection, to other gestures with similitude
Template carries out same detection step, then takes whole story point of the maximum segment of similarity measure as cutting in alternative collection, can will be defeated
Enter gesture and is converted into set metasequence.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610177623.5A CN105809144B (en) | 2016-03-24 | 2016-03-24 | A kind of gesture recognition system and method using movement cutting |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610177623.5A CN105809144B (en) | 2016-03-24 | 2016-03-24 | A kind of gesture recognition system and method using movement cutting |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105809144A CN105809144A (en) | 2016-07-27 |
| CN105809144B true CN105809144B (en) | 2019-03-08 |
Family
ID=56453890
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610177623.5A Active CN105809144B (en) | 2016-03-24 | 2016-03-24 | A kind of gesture recognition system and method using movement cutting |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105809144B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107909042B (en) * | 2017-11-21 | 2019-12-10 | 华南理工大学 | continuous gesture segmentation recognition method |
Families Citing this family (25)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108427194A (en) * | 2017-02-14 | 2018-08-21 | 深圳梦境视觉智能科技有限公司 | A kind of display methods and equipment based on augmented reality |
| CN107272878B (en) * | 2017-02-24 | 2020-06-16 | 广州幻境科技有限公司 | Identification method and device suitable for complex gesture |
| CN107092349A (en) * | 2017-03-20 | 2017-08-25 | 重庆邮电大学 | A kind of sign Language Recognition and method based on RealSense |
| CN106998409B (en) * | 2017-03-21 | 2020-11-27 | 华为技术有限公司 | Image processing method, head-mounted display and rendering equipment |
| CN106991398B (en) * | 2017-04-01 | 2020-03-27 | 北京工业大学 | Gesture recognition method based on image recognition and matched with graphical gloves |
| CN107092882B (en) * | 2017-04-19 | 2020-04-28 | 南京大学 | Behavior recognition system based on sub-action perception and working method thereof |
| CN108510525B (en) | 2018-03-30 | 2019-03-12 | 百度在线网络技术（北京）有限公司 | Template method for tracing, device, augmented reality system and storage medium |
| CN108596150A (en) * | 2018-05-10 | 2018-09-28 | 南京大学 | A kind of Activity recognition system and its working method excluding abnormal operation |
| CN109101879B (en) * | 2018-06-29 | 2022-07-01 | 温州大学 | Posture interaction system for VR virtual classroom teaching and implementation method |
| CN110472396B (en) * | 2018-08-17 | 2022-12-30 | 中山叶浪智能科技有限责任公司 | Somatosensory gesture touch method, system, platform and storage medium |
| CN110866864B (en) * | 2018-08-27 | 2024-10-22 | 阿里巴巴华东有限公司 | Face pose estimation/three-dimensional face reconstruction method and device and electronic equipment |
| CN110913279B (en) * | 2018-09-18 | 2022-11-01 | 中科海微(北京)科技有限公司 | Processing method for augmented reality and augmented reality terminal |
| CN109598229B (en) * | 2018-11-30 | 2024-06-21 | 李刚毅 | Monitoring system and method based on action recognition |
| CN109751998A (en) * | 2019-01-14 | 2019-05-14 | 重庆邮电大学 | A kind of recognizing model of movement method based on dynamic time warping |
| CN109966739B (en) * | 2019-01-17 | 2022-11-11 | 珠海金山数字网络科技有限公司 | Method and system for optimizing game operation |
| CN109977890B (en) * | 2019-03-30 | 2021-08-17 | 绵阳硅基智能科技有限公司 | Action recognition method and recognition system thereof |
| CN110163086B (en) * | 2019-04-09 | 2021-07-09 | 有品国际科技(深圳)有限责任公司 | Body-building action recognition method, device, equipment and medium based on neural network |
| CN110163113B (en) * | 2019-04-25 | 2023-04-07 | 上海师范大学 | Human behavior similarity calculation method and device |
| CN110059661B (en) * | 2019-04-26 | 2022-11-22 | 腾讯科技（深圳）有限公司 | Action recognition method, man-machine interaction method, device and storage medium |
| CN110263743B (en) * | 2019-06-26 | 2023-10-13 | 北京字节跳动网络技术有限公司 | Method and device for recognizing images |
| CN110263742A (en) * | 2019-06-26 | 2019-09-20 | 北京字节跳动网络技术有限公司 | The method and apparatus of image for identification |
| CN112181148A (en) * | 2020-09-29 | 2021-01-05 | 中国人民解放军军事科学院国防科技创新研究院 | Multimodal man-machine interaction method based on reinforcement learning |
| CN112464847B (en) * | 2020-12-07 | 2021-08-31 | 北京邮电大学 | Human body action segmentation method and device in video |
| CN113840177B (en) * | 2021-09-22 | 2024-04-30 | 广州博冠信息科技有限公司 | Live interaction method and device, storage medium and electronic equipment |
| CN114533039B (en) * | 2021-12-27 | 2023-07-25 | 重庆邮电大学 | Human joint position and angle resolving method based on redundant sensor |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102508547A (en) * | 2011-11-04 | 2012-06-20 | 哈尔滨工业大学深圳研究生院 | Computer-vision-based gesture input method construction method and system |
| CN102982315A (en) * | 2012-11-05 | 2013-03-20 | 中国科学院计算技术研究所 | Gesture segmentation recognition method capable of detecting non-gesture modes automatically and gesture segmentation recognition system |
| CN103649967A (en) * | 2011-06-23 | 2014-03-19 | 阿尔卡特朗讯 | Dynamic gesture recognition process and authoring system |
| CN104933408A (en) * | 2015-06-09 | 2015-09-23 | 深圳先进技术研究院 | Hand gesture recognition method and system |
| CN105320937A (en) * | 2015-09-25 | 2016-02-10 | 北京理工大学 | Kinect based traffic police gesture recognition method |
-
2016
- 2016-03-24 CN CN201610177623.5A patent/CN105809144B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103649967A (en) * | 2011-06-23 | 2014-03-19 | 阿尔卡特朗讯 | Dynamic gesture recognition process and authoring system |
| CN102508547A (en) * | 2011-11-04 | 2012-06-20 | 哈尔滨工业大学深圳研究生院 | Computer-vision-based gesture input method construction method and system |
| CN102982315A (en) * | 2012-11-05 | 2013-03-20 | 中国科学院计算技术研究所 | Gesture segmentation recognition method capable of detecting non-gesture modes automatically and gesture segmentation recognition system |
| CN104933408A (en) * | 2015-06-09 | 2015-09-23 | 深圳先进技术研究院 | Hand gesture recognition method and system |
| CN105320937A (en) * | 2015-09-25 | 2016-02-10 | 北京理工大学 | Kinect based traffic police gesture recognition method |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107909042B (en) * | 2017-11-21 | 2019-12-10 | 华南理工大学 | continuous gesture segmentation recognition method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105809144A (en) | 2016-07-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105809144B (en) | A kind of gesture recognition system and method using movement cutting | |
| TWI430185B (en) | Facial expression recognition systems and methods and computer program products thereof | |
| Li et al. | Joint distance maps based action recognition with convolutional neural networks | |
| Shreve et al. | Macro-and micro-expression spotting in long videos using spatio-temporal strain | |
| US8401248B1 (en) | Method and system for measuring emotional and attentional response to dynamic digital media content | |
| CN102375970B (en) | A kind of identity identifying method based on face and authenticate device | |
| CN105740779B (en) | Method and device for detecting living human face | |
| JP3962803B2 (en) | Head detection device, head detection method, and head detection program | |
| CN107092349A (en) | A kind of sign Language Recognition and method based on RealSense | |
| CN107480586B (en) | Face characteristic point displacement-based biometric photo counterfeit attack detection method | |
| Marcos-Ramiro et al. | Body communicative cue extraction for conversational analysis | |
| CN110796101A (en) | Face recognition method and system of embedded platform | |
| CN104821010A (en) | Binocular-vision-based real-time extraction method and system for three-dimensional hand information | |
| Monir et al. | Rotation and scale invariant posture recognition using Microsoft Kinect skeletal tracking feature | |
| US11676357B2 (en) | Modification of projected structured light based on identified points within captured image | |
| CN104794446B (en) | Human motion recognition method and system based on synthesis description | |
| CN106599779A (en) | Human ear recognition method | |
| Hoque et al. | Computer vision based gesture recognition for desktop object manipulation | |
| Zraqou et al. | An efficient approach for recognizing and tracking spontaneous facial expressions | |
| CN104751144B (en) | A kind of front face fast appraisement method of facing video monitoring | |
| CN110705453A (en) | Real-time fatigue driving detection method | |
| Liang et al. | Face pose estimation using near-infrared images | |
| CN111652019B (en) | Face living body detection method and device | |
| KR20130081126A (en) | Method for hand-gesture recognition and apparatus thereof | |
| Afroze et al. | Talking vs non-talking: A vision based approach to detect human speaking mode |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |