CN106598234A - Gesture recognition method based on inertial sensing - Google Patents

Gesture recognition method based on inertial sensing Download PDF

Info

Publication number
CN106598234A
CN106598234A CN201611072669.7A CN201611072669A CN106598234A CN 106598234 A CN106598234 A CN 106598234A CN 201611072669 A CN201611072669 A CN 201611072669A CN 106598234 A CN106598234 A CN 106598234A
Authority
CN
China
Prior art keywords
gesture
data
main shaft
sigma
sample
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN201611072669.7A
Other languages
Chinese (zh)
Other versions
CN106598234B (en
Inventor
夏侯士戟
王琳琳
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
University of Electronic Science and Technology of China
Original Assignee
University of Electronic Science and Technology of China
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by University of Electronic Science and Technology of China filed Critical University of Electronic Science and Technology of China
Priority to CN201611072669.7A priority Critical patent/CN106598234B/en
Publication of CN106598234A publication Critical patent/CN106598234A/en
Application granted granted Critical
Publication of CN106598234B publication Critical patent/CN106598234B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • 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
    • G06F3/017—Gesture based interaction, e.g. based on a set of recognized hand gestures
    • 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
    • G06F3/03—Arrangements for converting the position or the displacement of a member into a coded form
    • G06F3/033—Pointing devices displaced or positioned by the user, e.g. mice, trackballs, pens or joysticks; Accessories therefor
    • G06F3/0346—Pointing devices displaced or positioned by the user, e.g. mice, trackballs, pens or joysticks; Accessories therefor with detection of the device orientation or free movement in a 3D space, e.g. 3D mice, 6-DOF [six degrees of freedom] pointers using gyroscopes, accelerometers or tilt-sensors

Landscapes

  • Engineering & Computer Science (AREA)
  • General Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Human Computer Interaction (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses a gesture recognition method based on inertial sensing. The method comprises the following steps: collecting gesture data by using a three-axis inertial sensor, performing main axis judgment on each gesture action in a gesture data sample library obtained by collecting, separately clustering the gesture data of each experimenter under each gesture action in the gesture data sample library, screening out typical samples to construct a typical sample set of the gesture action, during gesture recognition, firstly performing main axis judgment on the gesture data of the tested gesture action, then screening out the gesture actions of which the main axis is the same as the main axis of the tested gesture action from the gesture actions in the gesture data sample library, calculating the similarity between the tested gesture data and the typical sample set of the gesture actions, and selecting the gesture action having the maximum similarity as the category to which the tested gesture belongs. According to the gesture recognition method disclosed by the invention, by means of the main axis judgment and the screening of the typical samples, the complexity of the algorithm can be effectively reduced, and the gesture recognition accuracy is improved.

Description

Gesture identification method based on inertia sensing
Technical field
The invention belongs to technical field of hand gesture recognition, more specifically, is related to a kind of gesture based on inertia sensing and knows Other method.
Background technology
Recently as technology development, human-computer interaction technology has obtained relatively broad application, human body hand in life Gesture can transmit significant information with more natural as most abundant, most significant human action to environment, effectively reduce learning Practise cost.Due to the individual gesture motion of different user it is larger in the many-side differentiation such as speed, amplitude, based on inertia sensing Gesture is followed the trail of and the gesture identification method for how making not relying on individuality is concentrated on the research focus of recognition methods with more individuality Robustness, while obtaining faster dynamic response.
In existing academic research, Ruize Xu et al. propose that a kind of gesture based on gesture mark and template matching algorithm is known Other method, in the gesture identification problem for not relying on individuality higher discrimination is obtained.But it is suitable to fairly simple gesture Action, for complicated gesture motion, mark is difficult.Kuang-Yow Lian et al. propose that a kind of feature based is extracted and trained The gesture identification method of hidden Markov model is improved, but due to user's individual difference, there is larger difference in same hand signal, It is difficult to set up accurate gesture template and hidden Markov model.K Barczewska et al. compared for three kinds it is different based on The gesture identification method (DTW, DDTW, PDTW) of DTW (Dynamic Time Warping, dynamic time consolidation) algorithm it is excellent It is bad, test result indicate that：DDTW methods enhance algorithm and signal local are become to carrying out DTW computings again after signal derivation process The adaptability of change, with highest recognition accuracy；The recognition accuracy of PDTW methods is minimum, but for the place of mass data collection Reason can effectively shorten operation time, the operation time of different compression ratio determining method.Hussain SMA et al. are for connecting Continuous gesture recognition system, with acceleration transducer multidimensional data DTW calculating is carried out, and algorithm identification is improve to a certain extent Rate, but increased operation time.
Above several method is randomly selected out the sample that a part is matched as DTW algorithm templates by the data of collection, Specially treated is not done to sample set, it is impossible to ensure standard degree, validity and the availability of sample.
It is " CN105824420A ", entitled " a kind of gesture identification method based on acceleration transducer " in publication No. Disclose a kind of gesture identification method in patent, the method lay particular emphasis on can automatic decision action beginning and end, it is not necessary to outward The control intervention on boundary.Non-compliant gesture is excluded by three steps (angle, feature, state) step, overall calculating is reduced Amount.But it is poor for the pardon of Different Individual feature in experiment because action planting modes on sink characteristic is obtained by a large amount of gesture samples of collection, Gesture identification accuracy rate is not high.
The content of the invention
It is an object of the invention to overcome the deficiencies in the prior art, there is provided a kind of gesture identification side based on inertia sensing Method, by judging and typical sample screening using main shaft, can effectively reduce the complexity of algorithm, improve gesture identification accurate Rate.
For achieving the above object, the present invention is comprised the following steps based on the gesture identification method of inertia sensing：
S1：Q gesture motion data for gathering M position experimenters using three axle inertial sensors build gesture data sample Storehouse, each experimenter repeats n times to each gesture motion, remembers three of m position experimenters to q-th gesture motion n-th repetition Axle gesture data is Gq,m,n, wherein q=1,2 ..., Q, m=1,2 ..., M, n=1,2 ..., N；
S2：Main shaft judgement is carried out to each gesture motion, from three axles main shaft is filtered out；
S3：The n times gesture data sample set for remembering each experimenter under each gesture motion is Gq,m={ Gq,m,1,…, Gq,m,N, respectively to each sample set Gq,mClustered using the indefinite clustering algorithm of cluster numbers, the cluster numbers for obtaining are designated as Kq,m, in each cluster the nearest gesture data sample of chosen distance cluster centre as typical sample, so as to obtain each hand The typical sample collection Y of gesture actionq={ Yq,1,Yq,2,…,Yq,M, wherein, Represent q K-th typical sample of m-th experimenter, k=1,2 ..., K under individual gesture motionq,m；
S4：Using the gesture data C of certain test gesture motion of three axle inertial sensor collecting test persons；
S5：Main shaft judgement is carried out to testing gesture motion according to test gesture motion data C, from three axles master is filtered out Axle；
S6：Filter out from Q gesture motion of gesture data Sample Storehouse and test gesture motion main shaft identical P hand Gesture action, then calculates respectively the typical sample collection Y of test gesture motion data C and P gesture motionpSimilarity D (C, Yp)：
Wherein, p=1,2 ..., P,Represent test gesture motion data C and typical sample collection YpMiddle typical case SampleBetween similarity, then select D (C, Yp) gesture motion corresponding to minimum of a value is used as the test gesture motion Generic, complete gesture identification.
Gesture identification method of the present invention based on inertia sensing, using three axle inertial sensors gesture data collection is carried out, Main shaft judgement is carried out to each gesture motion in the gesture data Sample Storehouse that collects, then in gesture data sample storehouse Each gesture motion under the gesture data of each experimenter clustered respectively, filter out typical sample and build the gesture motion Typical sample collection, in gesture identification, first to test gesture motion gesture data carry out main shaft judgement, then from gesture Main shaft identical gesture motion is screened in the gesture motion in data sample storehouse, test gesture data is then calculated and is moved with these gestures Make the similarity between typical sample collection, select the maximum gesture motion of similarity as the generic of test gesture.
The present invention has following technique effect：
(1) for gesture data Sample Storehouse, carry out respectively for the gesture data of each experimenter under each gesture motion Cluster, filters out typical sample, can improve the normality and validity of sample dictionary, so as to improve the accurate of gesture identification Rate；
(2) concentrated position of gesture data useful information is gone out using main shaft determination methods Effective selection, so as to know in gesture Gesture motion initial screening is carried out by main shaft when other, so as to effectively reduce the complexity of algorithm, gesture identification accuracy rate is improved With the robustness of individual consumer.
Description of the drawings
Fig. 1 is specific embodiment flow chart of the present invention based on the gesture identification method of inertia sensing；
Fig. 2 is gesture motion definition figure in the present embodiment.
Specific embodiment
The specific embodiment of the present invention is described below in conjunction with the accompanying drawings, so as to those skilled in the art preferably Understand the present invention.Requiring particular attention is that, in the following description, when known function and design detailed description perhaps When can desalinate the main contents of the present invention, these descriptions will be ignored here.
Fig. 1 is specific embodiment flow chart of the present invention based on the gesture identification method of inertia sensing.As shown in figure 1, The present invention is comprised the following steps based on the gesture identification method of inertia sensing：
S101：Collection gesture data Sample Storehouse：
Q gesture motion data for gathering M position experimenters using three axle inertial sensors build gesture data Sample Storehouse, often Individual experimenter repeats n times to each gesture motion, remembers three axle gestures of the m position experimenters to q-th gesture motion n-th repetition Data are Gq,m,n, wherein q=1,2 ..., Q, m=1,2 ..., M, n=1,2 ..., N.Obvious Gq,m,nFor xq,m,n× 3 matrix, xq,m,nRepresent Gq,m,nIn three number of axle evidences length, because the speed difference of individual operations causes the not same of same gesture action This time domain length disunity, xq,m,nSize also can be different.
In order to improve the accuracy and availability of gesture data, in general need to the gesture in gesture data sample storehouse Data are pre-processed, including but not limited to debounce, denoising, error compensation etc., and these preprocess methods are the normal of data processing With method, detailed process will not be described here.
S102：Main shaft judges：
Due to carrying out gesture data collection using three axle inertial sensors in the present invention, different gesture motions are in x, y, z Tendency on axle is different, that is to say, that possible data fluctuation can be very big on certain axle for some gesture motions, and another Almost may not fluctuate on outer axle, it is clear that the bigger axle of data fluctuations, comprising gesture data signal can be more, can regard It is the core information source of gesture data.Based on this, the present invention carries out main shaft judgement to each gesture motion, sieves from three axles Main shaft is selected, preliminary screening is carried out to gesture data according to main shaft in gesture identification, so as to reduce comparison during gesture identification Data volume, to improve the efficiency and accuracy rate of gesture identification.
The main shaft determination methods adopted in the present invention for：For each gesture motion, by all gesture data G under itq,m,n The gesture data vector of x, y, z axle be designated asNumber in this three axles gesture data vector is calculated respectively According to variance beThen variance of each gesture motion in x, y, z axle is averagely obtainedComputing formula is：
Screening varianceThe corresponding axle of middle maximum as main shaft, for other two axle, if its variance Predetermined threshold value is less than with the difference of variance maximum, then the axle is also served as into main shaft, otherwise not as main shaft.It can be seen that, gesture is moved It is main shaft at least to have an axle, and possible three axles are main shaft when most.
S103：Screening typical sample：
Due to there is great amount of samples in gesture data Sample Storehouse, the big gauge of needs is compared one by one in gesture identification Calculate, therefore need to select typical sample in all gesture data samples of each gesture motion in the present invention, for gesture Reference template during identification.Due to the otherness of gesture motion between individuality, the whether enough standards of sample form can directly affect hand The accuracy rate of gesture identification, it is clear that the mode for randomly selecting gesture data as typical sample is it cannot be guaranteed that the standard degree of sample set. Therefore the present invention selects typical sample by the way of cluster from the gesture data sample of each experimenter in each gesture motion This, to ensure standard degree, validity and the availability of typical sample, its concrete grammar is：
The n times gesture data sample set for remembering each experimenter under each gesture motion is Gq,m={ Gq,m,1,…,Gq,m,N, Respectively to each sample set Gq,mClustered using the indefinite clustering algorithm of cluster numbers, the cluster numbers for obtaining are designated as into Kq,m, The nearest gesture data sample of chosen distance cluster centre is used as typical sample in each cluster, it is clear that the quantity of typical sample For Kq,m, so as to obtain the typical sample collection Y of each gesture motionq={ Yq,1,Yq,2,…,Yq,M, wherein, Represent k-th typical sample of m-th experimenter under q-th gesture motion, k=1, 2,…,Kq,m.Obvious typical sample collection YqIn sample size be
In the present embodiment, using the semi-supervised clustering that (affinity propagation, AP) algorithm is propagated based on neighbour Method is respectively to each sample set Gq,mClustered.Novel semi-supervised based on neighbour's propagation algorithm need not refer in advance Determine clusters number, it all as potential cluster centre, enters all of data point on the basis of data point similarity matrix Row cluster, its objective is the representative point set for finding optimum so that similarity of all data points to nearest representative point Sum is maximum.It is different from other clustering algorithms, it is in esse point in sample set as the representative point of cluster centre, in AP Under clustering algorithm, can be directly using representative point as all kinds of typical samples.Therefore AP clustering algorithms are selected as gesture number According to the clustering method of sample, the selection that cluster result is limited to initial representative point can be prevented effectively from, make cluster result more Accurately.The concrete principle and step of AP clustering algorithms may refer to document " Xiao Yu；Yu Jian；Half prison based on neighbour's propagation algorithm Superintend and direct cluster [J]；Journal of Software；11 phases in 2008 ".
In cluster process, need to calculate the similarity between gesture data sample, because gesture data sample is three axles Gesture data, needs carry out Similarity Measure to each axle gesture data respectively, then averagely obtain between gesture data sample Similarity.Due to having carried out main shaft judgement to each gesture motion in the present invention, the main shaft with core information is filtered out, because This only can respectively be calculated main shaft phase when the similarity between two gesture data samples is calculated to main shaft gesture data Like spending, then the similarity between gesture data sample is averagely obtained to main shaft similarity.For example, it is assumed that certain gesture motion Main shaft is x-axis and z-axis, then for the gesture data sample two-by-two under it, it is only necessary to calculate x-axis data and z-axis data respectively Similarity, then sue for peace divided by 2.
Due to each gesture data Gq,m,nWill not be completely isometric, thus in the present embodiment preferred DTW algorithms calculating each axle The similarity of gesture data.DTW algorithms are a kind of non-linear consolidation skills that time consolidation and distance measurement calculations incorporated are got up Art, can preferably solve the uneven difficult problem of individual actions speed in gesture identification.DTW algorithms use certain specified attribute Non-linear warping function is allowed to another to the fluctuation approximate modeling on time shaft by bending the time shaft of one of pattern Individual pattern reaches farthest overlap, it is ensured that residual error distance is minimum, so as to eliminate the time between two space-time intermediate schemes Difference.The concrete principle and step of DTW algorithms may refer to document " Keogh E J, Pazzani M J.Derivative Dynamic Time Warping[C]//Sdm.2001,1:5-7.”。
S104：Collecting test gesture data：
Using the gesture data C of certain test gesture motion of three axle inertial sensor collecting test persons.
S105：Main shaft judges：
Main shaft judgement is carried out to testing gesture motion according to test gesture motion data C, from three axles main shaft is filtered out.With Step S102 is similar to, and its concrete grammar is：The gesture data vector of the x, y, z axle of note test gesture motion data T is Cx、Cy、 Cz, the vectorial variance of this three axles gesture data is calculated respectively is Screening varianceMiddle maximum Corresponding axle as main shaft, for other two axle, if its variance is less than predetermined threshold value with the difference of variance maximum, should Axle also serves as main shaft, otherwise not as main shaft.
S106：Gesture identification：
The main shaft of the test gesture motion for being obtained according to the judgement of step S105 first, from Q hand of gesture data Sample Storehouse Main shaft identical P gesture motion is filtered out in gesture action, gesture data C and P of test gesture motion is then calculated respectively The typical sample collection Y of gesture motionpSimilarity D (C, Yp)：
Wherein, p=1,2 ..., P,Represent test gesture motion data C and typical sample collection YpMiddle typical case SampleBetween similarity.That is, first calculating the gesture data C and typical sample collection Y of test gesture motion respectivelyp In each typical sampleBetween similarity, then by gesture data C and typical sample collection YpIn all typical samples Similarity carry out averagely, you can obtain gesture data C and typical sample collection YpSimilarity.
Then D (C, Y are selectedp) gesture motion corresponding to minimum of a value as the test gesture motion generic, it is complete Into gesture identification.
Similarly, due to having carried out main shaft judgement in the present invention, and it is identical that main shaft has first been filtered out in gesture identification Gesture motion, therefore during gesture identification when gesture data and the typical sample similarity of test gesture motion is calculated, Only main shaft similarity is calculated respectively to main shaft gesture data, then main shaft similarity is averagely obtained gesture data sample it Between similarity.
Similarly, due to there is Length discrepancy between the gesture data and typical sample of testing gesture motion, therefore When the gesture data of test gesture motion is calculated with typical sample similarity, Similarity Measure is carried out using DTW algorithms.
Embodiment
In order to illustrate the technique effect of the present invention, experimental verification is carried out using a specific example, and calculated from DTW Method, DDTW algorithms and PDTW algorithms algorithm as a comparison, contrasts to experiment effect.Fig. 2 is gesture motion in the present embodiment Definition figure.As shown in Fig. 2 in this experimental verification, defining 12 gesture motions, complete in x-z-plane, solid black Circled is gesture starting point, and dotted line is that hand returns initial point route after the completion of gesture motion.This experiment acquire altogether from The gesture data of 10 volunteers (5 male 5 female), with up-down-right-left-circle_1-circle_2-tick_1- Tick_2-tick down_1-tick down_2-slant up-slant down orders are set, it is desirable to per aspiration Person completes the collection of 20 groups of action datas according to the gesture motion that Fig. 2 is defined, altogether 2400 gesture datas.Select different gestures 960 gesture datas that amount to of action constitute gesture data Sample Storehouse.Then carry out main shaft respectively to this 12 gesture motions to sentence Disconnected, table 1 is the main shaft judged result of gesture motion in gesture data Sample Storehouse in the present embodiment.
Gesture Main shaft Gesture Main shaft
up Z axis tick_1 X-axis, Z axis
down Z axis tick_2 X-axis, Z axis
right X-axis tick down_1 X-axis, Z axis
left X-axis tick down_2 X-axis, Z axis
circle_1 X-axis, Z axis slant up X-axis, Z axis
circle_2 X-axis, Z axis slant down X-axis, Z axis
Table 1
Filtered out in gesture data Sample Storehouse after the typical sample of each gesture motion, using remaining using clustering algorithm 1440 gesture datas carry out gesture identification as the gesture data of test gesture motion, and statistics obtains the knowledge of different gesture motions Other accuracy rate.Table 2 is of the invention for the different gesture identification degrees of accuracy (%) for depending on individuality, and table 3 is the present invention for not Depend on the different gesture identification degrees of accuracy (%) of individuality.
Table 2
Table 3
Visible according to table 2 and table 3, for different gesture motions, there is some difference for the recognition accuracy of the present invention, its In with slant down most easily to obscure gesture motion.Knowable to being counted to all gesture motions, the present invention is for depending on The recognition accuracy of individual gesture identification is up to 98.84%；For the recognition accuracy of the gesture identification for not relying on individuality Up to 96.7%.Table 4 is the recognition accuracy contrast of the present invention and three kinds of contrast algorithms.
Algorithm Depend on individuality Individuality is not relied on
DTW 86.2% 74.9%
DDTW 91.2% 78.9%
PDTW 85.8% 69.8%
The present invention 98.84% 96.7%
Table 4
It can be seen from table 4, the gesture identification that individuality is also not dependent on individuality is to rely on without recognizing, the identification of the present invention is accurate Really rate is above three kinds of contrast algorithms, and being especially not dependent on the recognition accuracy in terms of individuality has significant raising.
Although being described to illustrative specific embodiment of the invention above, in order to the technology of the art Personnel understand the present invention, it should be apparent that the invention is not restricted to the scope of specific embodiment, the common skill to the art For art personnel, as long as various change is in the spirit and scope of the present invention of appended claim restriction and determination, these Change is it will be apparent that all utilize the innovation and creation of present inventive concept in the row of protection.

Claims (5)

1. a kind of gesture identification method based on inertia sensing, it is characterised in that comprise the following steps：
S1：Q gesture motion data for gathering M position experimenters using three axle inertial sensors build gesture data Sample Storehouse, often Individual experimenter repeats n times to each gesture motion, remembers three axle gestures of the m position experimenters to q-th gesture motion n-th repetition Data are Gq,m,n, wherein q=1,2 ..., Q, m=1,2 ..., M, n=1,2 ..., N；
S2：Main shaft judgement is carried out to each gesture motion, from three axles main shaft is filtered out；
S3：The n times gesture data sample set for remembering each experimenter under each gesture motion is Gq,m={ Gq,m,1,…,Gq,m,N, point It is other to each sample set Gq,mClustered using the indefinite clustering algorithm of cluster numbers, the cluster numbers for obtaining are designated as into Kq,m, every The nearest gesture data sample of chosen distance cluster centre is used as typical sample in individual cluster, so as to obtain each gesture motion Typical sample collection Yq={ Yq,1,Yq,2,…,Yq,M, wherein, Represent that q-th gesture is moved Make k-th typical sample of lower m-th experimenter, k=1,2 ..., Kq,m；
S4：Using the gesture data C of certain test gesture motion of three axle inertial sensor collecting test persons；
S5：Main shaft judgement is carried out to testing gesture motion according to test gesture motion data C, from three axles main shaft is filtered out；
S6：Filter out from Q gesture motion of gesture data Sample Storehouse and moved with test gesture motion main shaft identical P gesture Make, the typical sample collection Y of test gesture motion data C and P gesture motion is then calculated respectivelypSimilarity D (C, Yp)：
D ( C , Y p ) = Σ m = 1 M Σ k = 1 K p , m D ( C , G p , m , k * ) Σ m = 1 M K p , m
Wherein, p=1,2 ..., P,Represent test gesture motion data G ' and typical sample collection YpMiddle typical sampleBetween similarity；
Then D (C, Y are selectedp) gesture motion corresponding to minimum of a value, as the generic of the test gesture motion, completes gesture Identification.
2. gesture identification method according to claim 1, it is characterised in that the method that main shaft judges in step S2 For：For each gesture motion, by all gesture data G under itq,m,nThe gesture data vector of x, y, z axle be designated asThe variance for calculating data in this three axles gesture data vector respectively is Then variance of each gesture motion in x, y, z axle is averagely obtainedComputing formula is：
σ q x = Σ m = 1 M Σ n = 1 N σ q , m , n x M × N
σ q y = Σ m = 1 M Σ n = 1 N σ q , m , n y M × N
σ q z = Σ m = 1 M Σ n = 1 N σ q , m , n z M × N
Screening varianceThe corresponding axle of middle maximum as main shaft, for other two axle, if its variance and side The difference of difference maximum is less than predetermined threshold value, then the axle is also served as into main shaft, otherwise not as main shaft；
The concrete grammar of main shaft judgement is in step S5：Note test gesture motion data T x, y, z axle gesture data to Measure as Cx、Cy、Cz, the vectorial variance of this three axles gesture data is calculated respectively is Screening varianceThe corresponding axle of middle maximum as main shaft, for other two axle, if the difference of its variance and variance maximum Value is less than predetermined threshold value, then the axle is also served as into main shaft, otherwise not as main shaft.
3. gesture identification method according to claim 1, it is characterised in that the clustering algorithm in step S3 adopts base In the Novel semi-supervised of neighbour's propagation algorithm.
4. gesture identification method according to claim 1, it is characterised in that to sample set G in step S3q,mCarry out Cluster process calculates the gesture data of test gesture motion and typical case in gesture data Sample Similarity and step S6 is calculated During Sample Similarity, only main shaft similarity is calculated respectively to main shaft gesture data, then main shaft similarity is averagely obtained Similarity between gesture data sample.
5. gesture identification method according to claim 1, it is characterised in that to sample set G in step S3q,mCarry out Cluster process calculates the gesture data of test gesture motion and typical case in gesture data Sample Similarity and step S6 is calculated During Sample Similarity, Similarity Measure is carried out using DTW algorithms.
CN201611072669.7A 2016-11-28 2016-11-28 Gesture identification method based on inertia sensing Expired - Fee Related CN106598234B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201611072669.7A CN106598234B (en) 2016-11-28 2016-11-28 Gesture identification method based on inertia sensing

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201611072669.7A CN106598234B (en) 2016-11-28 2016-11-28 Gesture identification method based on inertia sensing

Publications (2)

Publication Number Publication Date
CN106598234A true CN106598234A (en) 2017-04-26
CN106598234B CN106598234B (en) 2019-05-28

Family

ID=58594044

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201611072669.7A Expired - Fee Related CN106598234B (en) 2016-11-28 2016-11-28 Gesture identification method based on inertia sensing

Country Status (1)

Country Link
CN (1) CN106598234B (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109521877A (en) * 2018-11-08 2019-03-26 中国工商银行股份有限公司 Mobile terminal man-machine interaction method and system
CN110377159A (en) * 2019-07-24 2019-10-25 张洋 Action identification method and device
CN114625246A (en) * 2022-02-14 2022-06-14 深圳市心流科技有限公司 Gesture combination triggering method and device, intelligent bionic hand and storage medium
CN117075742A (en) * 2023-10-17 2023-11-17 深圳市魔样科技有限公司 Intelligent finger ring control method based on gesture recognition

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103984416A (en) * 2014-06-10 2014-08-13 北京邮电大学 Gesture recognition method based on acceleration sensor
CN105549746A (en) * 2016-01-28 2016-05-04 广州成潮智能科技有限公司 Action identification method based on acceleration sensing chip
US9383895B1 (en) * 2012-05-05 2016-07-05 F. Vinayak Methods and systems for interactively producing shapes in three-dimensional space
CN106096662A (en) * 2016-06-24 2016-11-09 深圳市颐通科技有限公司 Human motion state identification based on acceleration transducer

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9383895B1 (en) * 2012-05-05 2016-07-05 F. Vinayak Methods and systems for interactively producing shapes in three-dimensional space
CN103984416A (en) * 2014-06-10 2014-08-13 北京邮电大学 Gesture recognition method based on acceleration sensor
CN105549746A (en) * 2016-01-28 2016-05-04 广州成潮智能科技有限公司 Action identification method based on acceleration sensing chip
CN106096662A (en) * 2016-06-24 2016-11-09 深圳市颐通科技有限公司 Human motion state identification based on acceleration transducer

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109521877A (en) * 2018-11-08 2019-03-26 中国工商银行股份有限公司 Mobile terminal man-machine interaction method and system
CN110377159A (en) * 2019-07-24 2019-10-25 张洋 Action identification method and device
CN114625246A (en) * 2022-02-14 2022-06-14 深圳市心流科技有限公司 Gesture combination triggering method and device, intelligent bionic hand and storage medium
CN117075742A (en) * 2023-10-17 2023-11-17 深圳市魔样科技有限公司 Intelligent finger ring control method based on gesture recognition
CN117075742B (en) * 2023-10-17 2024-01-30 深圳市魔样科技有限公司 Intelligent finger ring control method based on gesture recognition

Also Published As

Publication number Publication date
CN106598234B (en) 2019-05-28

Similar Documents

Publication Publication Date Title
CN101694692B (en) Gesture identification method based on acceleration transducer
Batool et al. Telemonitoring of daily activity using accelerometer and gyroscope in smart home environments
Xu et al. A novel ensemble of random forest for assisting diagnosis of Parkinson's disease on small handwritten dynamics dataset
US10795481B2 (en) Method and system for identifying tap events on touch panel, and touch-controlled end product
CN106598234A (en) Gesture recognition method based on inertial sensing
CN103544392B (en) Medical science Gas Distinguishing Method based on degree of depth study
CN107122752B (en) Human body action comparison method and device
CN106897592A (en) User authen method, user authentication device and writing implement
CN105184325A (en) Human body action recognition method and mobile intelligent terminal
CN103116744A (en) Fake fingerprint detection method based on markov random field (MRF) and support vector machine-k nearest neighbor (SVM-KNN) classification
US11328211B2 (en) Delimitation in unsupervised classification of gestures
CN113176022B (en) Segmented neural network pressure sensor pressure detection method and system
CN103218825A (en) Quick detection method of spatio-temporal interest points with invariable scale
CN1373432A (en) Method and system for recognizing personal characteristics of scrip
Jin et al. Open-environment robotic acoustic perception for object recognition
Tan et al. Online signature verification based on dynamic features from gene expression programming
CN113705715B (en) Time sequence classification method based on LSTM and multi-scale FCN
CN110163130A (en) A kind of random forest grader and classification method of the feature pre-align for gesture identification
CN105678348A (en) Normative evaluation method and system for handwritten Chinese characters
Liu Error awareness by lower and upper bounds in ensemble learning
CN112380903A (en) Human activity identification method based on WiFi-CSI signal enhancement
CN109992106B (en) Gesture track recognition method, electronic device and storage medium
Gupta et al. S-pencil: A smart pencil grip monitoring system for kids using sensors
CN113609999B (en) Human body model building method based on gesture recognition
Shintani et al. Digital pen for handwritten alphabet recognition

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20190528

Termination date: 20211128

CF01 Termination of patent right due to non-payment of annual fee