CN106909891A - A kind of Human bodys' response method based on self feed back gene expression programming - Google Patents
A kind of Human bodys' response method based on self feed back gene expression programming Download PDFInfo
- Publication number
- CN106909891A CN106909891A CN201710059525.6A CN201710059525A CN106909891A CN 106909891 A CN106909891 A CN 106909891A CN 201710059525 A CN201710059525 A CN 201710059525A CN 106909891 A CN106909891 A CN 106909891A
- Authority
- CN
- China
- Prior art keywords
- human body
- artis
- motion
- modelling
- gene expression
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Psychiatry (AREA)
- Social Psychology (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of Human bodys' response method based on self feed back gene expression programming, the method is directed to human body behavior depth image, the three-dimensional time sequence data of human body multiple artis is therefrom extracted as sample, obtain self feed back gene expression programming sample is modeled using by gene expression programming adding TIS after intersection, mutation operation and inserting string operation construction, modelling of human body motion is obtained, TIS inserts string operation and refers to inserting functor string in artis motion sequence head；Then gradient information is extracted as the aspect of model.The aspect of model of the modelling of human body motion of training sample is input into neutral net, training obtains neural network model as human body behavior grader；The aspect of model of the corresponding modelling of human body motion of test sample is input into the human body behavior grader of above-mentioned acquisition, Human bodys' response result is obtained.The present invention has the advantages that the Human bodys' response degree of accuracy is high and recognition speed is fast.
Description
Technical field
The present invention relates to Human bodys' response field, more particularly to a kind of human body based on self feed back gene expression programming
Activity recognition method.
Background technology
In recent years, Human bodys' response research is one of most popular Some Questions To Be Researched of Computer Subject in the world, tool
There is important theoretical research value, research topic is abundant in content, be related to image processing and analysis, machine vision, Human Physiology
The multidisciplinary domain knowledge such as, human cinology, pattern-recognition and artificial intelligence.Human bodys' response research has not only widened
Research direction between section, and the development of related discipline is driven, while being also smart city in the urgent need to address with public safety
Key technology, it is widely used in human-computer intellectualization, intelligent vision monitoring, content based video retrieval system, virtual reality
Deng field.The research in the big portion of current human bioequivalence also only stops at the identification of static state, the complexity and mutability of human motion
So that the Accuracy and high efficiency of identification cannot meet the real requirement of relevant industries.
At present, the identification in the prior art to human body behavior is mainly using artis wearable sensors or using many
Camera carries out various visual angles monitoring, and artis wearable sensors refer to put sensor, this mode at certain position of human body
The behavior limitednumber of human body is recognized, and recognition accuracy is not high, recognition speed is slow, identification behavior is single and poor expandability；
It refers to install camera in each view position to carry out various visual angles monitoring using multi-cam, is clapped by the camera at each visual angle
The picture taken the photograph judges the behavior of human body, this mode exist high cost, data precision it is low, using inconvenience, by environmental factor shadow
Ring the defect such as big.
The content of the invention
Shortcoming and deficiency it is an object of the invention to overcome prior art, there is provided one kind is based on self feed back gene expression
The Human bodys' response method of programming, the method sets up modelling of human body motion using self feed back gene expression (SGEP), then
Each anthropoid behavior is accurately portrayed using the gradient information of modelling of human body motion as modelling of human body motion feature, is finally led to
Cross neutral net to be identified human body behavior, have the advantages that the Human bodys' response degree of accuracy is high and recognition speed is fast.
The purpose of the present invention is achieved through the following technical solutions：A kind of human body row based on self feed back gene expression programming
It is recognition methods, it is characterised in that step is as follows：
S1, acquisition human body behavior depth image, then extract the N number of joint of human body respectively from human body behavior depth image
Each self-corresponding three-dimensional time sequence data of point, as a sample；
S2, sample is modeled using self feed back gene expression programming, gets the corresponding each artis fortune of sample
Dynamic sequence, so as to obtain the modelling of human body motion that each sample is based on artis；Wherein described self feed back gene expression programming by
Gene expression programming adds the slotting string operation constructions of TIS after intersection, mutation operation and obtains, and wherein TIS inserts what string operation referred to
It is to insert functor string in artis motion sequence head, the length for increasing artis motion sequence significance bit increases；
S3, the gradient information of extraction modelling of human body motion are used as the aspect of model；
S4, training sample set is got by step S1, and each training sample is got respectively by way of step S2
The modelling of human body motion of training sample, then extracts the model of the modelling of human body motion of each training sample by way of step S3
Feature；
S5, the aspect of model of the modelling of human body motion of each training sample enter as the input of neutral net to neutral net
Row training, will train the neural network model for obtaining, as final human body behavior grader；
S6, test sample is got by step S1, and test sample gets test specimens by way of step S2
This modelling of human body motion, then extracts the aspect of model of the modelling of human body motion of test sample by way of step S3；
S7, the aspect of model of the modelling of human body motion of test sample is input into the human body behavior that is got into step S5 point
In class device, Human bodys' response result is obtained by the human body behavior grader.
Preferably, calculated using Microsoft the somatosensory device Kinect combination second generations SDK and OpenCV in the step S1
Machine vision storehouse obtains human body behavior depth image；TOF (Time Of Flight, flight time) technology is wherein utilized, is calculated micro-
The IR of infrared sensor transmitting reaches the phase difference of human body back reflection in soft Kinect sensor, obtains based on human body
Human body behavior depth image；
The corresponding three-dimensional time sequence data of each artis is to be with the infrared sensor of Microsoft somatosensory device Kinect
What the rectangular coordinate system in space that origin is set up was obtained, the rectangular coordinate system in space includes x-axis, y-axis and z-axis, and wherein x-axis is square
To for parallel to Microsoft somatosensory device Kinect horizontal directions, to the left, y-axis positive direction is perpendicular to Microsoft somatosensory device Kinect
Upwards, z-axis positive direction is Microsoft's somatosensory device Kinect directions to incline direction.
Preferably, N is more than 25 in the step S1.
Preferably, in the step S1 after human body behavior depth image is got, by human body behavior depth image
Human body is split with background, is then converted to skeleton tracing system, so as to obtain multiple artis three-dimensional time sequences of human body
Column data.
Preferably, self feed back gene expression programming is modeled to sample in the step S2, obtains based on artis
Modelling of human body motion detailed process it is as follows：
S21, determine modelling of human body motion needed for functor collection and terminal symbol collection and terminal symbol number, and set circulation
End condition；
S22, the random artis motion sequence initial population of establishment, are gene expression by the expression of each artis motion sequence
Tree；Artis motion sequence is made up of functor collection and terminal symbol collection, and terminal symbol is concentrated includes artis three-dimensional time sequence number
According to middle time t and the coordinate components of artis three-dimensional time sequence data；
S23, the adaptive value for calculating each artis motion sequence in population, then estimate model or absolute by relative error
Error model is estimated to each artis motion sequence in population to the adaptability of environment, qualified in assessment result
In the case of, a number of individuality is selected as male parent from population；And judge whether to meet loop termination condition, if full
Foot, then jump to S26, otherwise into step S24；
S24, the male parent selected according to step S23, gene expression volume is carried out to the artis motion sequence in population
The intersection of journey, mutation operation；
S25, the artis motion sequence to population carry out TIS and insert string operation, obtain population at individual of future generation, return to step
S23；
S26, output model, obtain the modelling of human body motion of the function expression represented by optimal chromosome, i.e. artis.
Further, functor collection F is in step S21：
F={ S, C, T, I, R, N, P, Q, E, A, G, L, B, D, F, H }
It is as follows that each element in wherein functor collection F represents implication：
S represents that SIN function Sin, C represent that cosine function cos, T represent that tan tan, I represent arcsin function
Asin, R represent that inverse cosine function acos, N represent that arctan function atan, P represent that power function pow, Q represent square root function
Sqrt, E represent that natural number is bottom exponential function exp, A represent that ABS function abs, G represent that sign function sign, L are represented
Natural logrithm function ln, B represent that denary logarithm function lg, D represent that hyperbolic sine function sinh, F represent hyperbolic cosine
Function cosh, H represent hyperbolic tangent function tanh；
Terminal symbol collection described in step S21 includes multidimensional terminating character and constant terminating character collection；
Wherein multidimensional terminating character T is：
T={ a, b, c, d, e, f, g, h, i }；
Each element in wherein multidimensional terminal symbol collection T represents each independent variable in multidimensional variable respectively, in inspection human body fortune
By in the actual value substitution modelling of human body motion corresponding to each variable during movable model；
Wherein constant terminal symbol collection C is：C={ 1000, -1000 }.
Preferably, it is characterised in that the process that TIS inserts string operation is as follows：
As artis motion sequence effective length m=1, i.e. the root position of artis motion sequence is a constant, this
When, it is the functor string for getting at random to be made a variation to [L/4] position since artis motion sequence root position；Wherein [L/4] table
Show L/4 round numbers part, L is the length of artis motion sequence；While other characters of artis motion sequence head are successively
After move, beyond head length character string delete；Wherein L is the length of artis motion sequence, and h is artis motion sequence head
The length in portion；
When artis motion sequence imitates length 1<m<During h, in last functor of artis motion sequence significance bit
The random functor string for getting of insertion below h-m, other characters behind artis motion sequence head insertion point are moved afterwards successively,
Character string beyond head length is deleted.
Preferably, in the step S3, for each artis Ji(i=1,2..., N), N be sample in artis it is total
Number, the modelling of human body motion remembered is zi=Fi(x,y,t)；The aspect of model that modelling of human body motion is extracted in step S3 is specific
Process is as follows：
Equidistantly taking k moment is designated as tj, j=1,2 ..., k；
Calculate the gradient information of each artisWithI=1,2 ..., N；J=1,2 ..., k；WhereinRepresent i-th artis in t respectivelyjMoment corresponding x, y-coordinate value；
By the above-mentioned each artis gradient information being calculatedWithI=1,2 ..., N；J=1,
2 ..., k is z as modelling of human body motioni=FiThe aspect of model of (x, y, t).
Further, trained by the aspect of model of the modelling of human body motion of each training sample in the step S5 and obtained
The detailed process of neural network model is as follows：
S51, neural networks with single hidden layer framework is initially set up, the modelling of human body motion of each training sample will be obtained in step S4
The aspect of modelI=1,2 ..., N；J=1,2 ..., k as neutral net input training sample
Collection, is provided with N × 2k input neutral net, and it is N × 4k+1 to set hidden neuron number；Included according to human body behavior
Classification number M to set output neuron number be that M, i.e. output vector are { c1,c2,…,cs,…,cM, wherein cs∈ R, and take
It is [0,1] that value is interval, works as cs=max { c1,c2,…,cMWhen, csBe taken as 1, i.e. neutral net output vector for 0,0 ...,
1 ..., 0 }, represent that human body behavior belongs to s class behaviors；
S52, to set neural network learning rate η be the random number between 0~1, and neural network transformation function f (x) is set
It is Sigmoid functions, learning rules are continuous perceptron learning rules, and neutral net is trained using error backpropagation algorithm,
Obtain neural network model.
Preferably, it is characterised in that in the step S5, the model for being directed to the modelling of human body motion of each training sample is special
Levy, as the input of neutral net, tear cross-validation method training open by ten and obtain neural network model, as final human body row
It is grader.
The present invention has the following advantages and effect relative to prior art：
(1) Human bodys' response method of the present invention is directed to the human body behavior depth image for getting, and therefrom extracts respectively
Human body multiple each self-corresponding three-dimensional time sequence datas of artis as a sample, then using self feed back gene expression
Programming is modeled to sample, obtains each sample and is based on the modelling of human body motion of artis, and extracts modelling of human body motion
Gradient information is used as the aspect of model.After getting training sample set, training sample is concentrated into the corresponding human motion of training sample
The aspect of model of model is input into neutral net, and training is obtained neural network model as human body behavior grader；Get
After test sample, the human body behavior that the aspect of model of the corresponding modelling of human body motion of test sample is input into above-mentioned acquisition is classified
In device, Human bodys' response result is obtained by the human body behavior grader.The present invention is using by gene table of the prior art
The slotting string operation constructions of TIS are added after intersection, mutation operation up to formula programming (GEP) and obtains self feed back gene expression programming pair
Sample is modeled, and wherein TIS inserts string operation and refers to inserting functor string in artis motion sequence head, is closed for increasing
The length of joint movements sequence significance bit, such that it is able to increase the uniformity of population at individual, make it more can quickly converge to
The globally optimal solution of problem, it is to avoid be absorbed in locally optimal solution.Therefore self feed back gene expression programming is used in the present invention to sample
Originally it is modeled the speed of the degree of accuracy and the identification that can effectively improve Human bodys' response.
(2) using the gradient information representative for specific behavior as the model in modelling of human body motion in the present invention
Feature, middle compared to existing technology to use global or local motion feature as the aspect of model, the aspect of model of the invention can
More accurately reflect human body behavior campaign, greatly reduce feature complexity, improve feature representativeness and recognition correct rate.
(3) present invention is obtained using Microsoft's somatosensory device Kinect combination second generation SDK and OpenCV computer visions storehouse
Human body behavior depth image is taken, data precision can be effectively improved, solve to be imaged using tradition in the prior art to a certain extent
Equipment recognition efficiency is low, recognition effect by such environmental effects it is big and wear joint sensors high cost, in-convenience in use waiting
Technical problem；
Brief description of the drawings
Fig. 1 is the inventive method flow chart.
Fig. 2 is that self feed back gene expression programming sets up modelling of human body motion flow chart in the inventive method.
Fig. 3 is that self feed back gene expression programming sets up the expression set up during modelling of human body motion in the inventive method
Formula tree graph.
Fig. 4 is expression tree map generalization process in Fig. 3.
Specific embodiment
With reference to embodiment and accompanying drawing, the present invention is described in further detail, but embodiments of the present invention are not limited
In this.
Embodiment
Present embodiment discloses a kind of Human bodys' response method based on self feed back gene expression programming, its feature exists
In as shown in figure 1, step is as follows：
S1, acquisition human body behavior depth image, then extract the N number of joint of human body respectively from human body behavior depth image
Each self-corresponding three-dimensional time sequence data of point, the three-dimensional of the N number of artis of human body is included as the sample of a sample, i.e.,
Time series data；Wherein counted using Microsoft the somatosensory device Kinect combination second generations SDK and OpenCV in the present embodiment
Calculation machine vision storehouse obtains human body behavior depth image, using TOF (Time Of Flight, flight time) technology, calculates Microsoft
The IR of infrared sensor transmitting reaches the phase difference of human body back reflection in Kinect sensor, obtains based on artis
Human body behavior depth image.The corresponding three-dimensional time sequence data of each artis is with the infrared of Microsoft somatosensory device Kinect
Sensor is obtained by rectangular coordinate system in space that origin is set up, and the rectangular coordinate system in space includes x-axis, y-axis and z-axis, its
Middle x-axis positive direction be parallel to Microsoft somatosensory device Kinect horizontal directions to the left, y-axis positive direction is to be set perpendicular to Microsoft's body-sensing
Upwards, z-axis positive direction is Microsoft's somatosensory device Kinect directions to standby Kinect incline directions.People is being got in this step
After body behavior depth image, the human body in human body behavior depth image is split with background, be then converted to skeleton tracking
System, so as to obtain multiple artis three-dimensional time sequence datas of human body, N is 25 in the present embodiment, i.e., from human body behavior depth
The corresponding three-dimensional time sequence data of 25 artis of human body is extracted respectively in degree image.
S2, sample is modeled using self feed back gene expression programming, gets the corresponding each artis fortune of sample
Dynamic sequence, so as to obtain the modelling of human body motion that each sample is based on artis；Wherein described self feed back gene expression programming
(SGEP) add the slotting string operation constructions of TIS after intersection, mutation operation by gene expression programming (GEP) to obtain, wherein TIS
Slotting string operation refers to inserting functor string in artis motion sequence head, for increasing artis motion sequence significance bit
Length increases.
S3, the gradient information of extraction modelling of human body motion are used as the aspect of model；For each artis Ji(i=1,2...,
N), the modelling of human body motion remembered is zi=Fi(x,y,t)；The aspect of model that modelling of human body motion is extracted in step S3 is specific
Process is as follows：
Equidistantly taking k moment is designated as tj, j=1,2 ..., k；
Calculate the gradient information of each artisWithI=1,2 ..., N；J=1,2 ..., k；WhereinRepresent i-th artis in t respectivelyjMoment corresponding x, y-coordinate value；
By the above-mentioned each artis gradient information being calculatedWithI=1,2 ..., N；J=1,
2 ..., k is z as modelling of human body motioni=FiThe aspect of model of (x, y, t).
S4, training sample set is got by step S1, and each training sample is got respectively by way of step S2
The modelling of human body motion of training sample, then extracts the model of the modelling of human body motion of each training sample by way of step S3
Feature.
S5, using the aspect of model of the modelling of human body motion of each training sample as neutral net input, by set god
Through the number of plies of network, the number of neuron, initialization weighting parameter and selection gradient optimization algorithm, neutral net is entered
Row training, neural network model as final human body behavior grader is obtained using training；Detailed process is as follows：
S51, neural networks with single hidden layer framework is initially set up, the modelling of human body motion of each training sample will be obtained in step S4
The aspect of modelI=1,2 ..., N；J=1,2 ..., k are input into neutral net, be provided with N ×
2k input neutral net, it is N × 4k+1 to set hidden neuron number；Set according to the classification sum M that human body behavior includes
It is that M, i.e. output vector are { c to put output neuron number1,c2,…,cs,…,cM, wherein cs∈ R, 1,2 ..., s ..., M point
Not Biao Shi human body behavior generic, and interval be [0,1], work as cs=max { c1,c2,…,cMWhen, cs1 is taken as, i.e. god
It is { 0,0 ..., 1 ..., 0 } through network output vector, represents that human body behavior belongs to s class behaviors；
S52, to set neural network learning rate η be the random number between 0~1, and neural network transformation function f (x) is set
Be Sigmoid functions, i.e. f (x)=1/ (1+e-x), learning rules are continuous perceptron learning rules, using error back propagation
Algorithm for Training neutral net, obtains neural network model.
In this step, for each training sample modelling of human body motion the aspect of model, as the input of neutral net,
Cross-validation method training is torn open by ten and obtain neural network model, as final human body behavior grader.
S6, test sample is got by step S1, and test sample gets test specimens by way of step S2
This modelling of human body motion, then extracts the aspect of model of the modelling of human body motion of test sample by way of step S3.
S7, the aspect of model of the modelling of human body motion of test sample is input into the human body behavior that is got into step S5 point
In class device, Human bodys' response result is obtained by the human body behavior grader.
In the present embodiment above-mentioned steps S2, as shown in Fig. 2 self feed back gene expression programming is modeled to sample, obtain
The detailed process that the modelling of human body motion of artis is based on to each sample is as follows：
S21, determine modelling of human body motion needed for functor collection and terminal symbol collection and terminal symbol number, and set circulation
End condition, subsequently into step S22；Functor collection F is wherein in this step：
F={ S, C, T, I, R, N, P, Q, E, A, G, L, B, D, F, H }
It is as follows that each element in wherein functor collection F represents implication：
S represents that SIN function Sin, C represent that cosine function cos, T represent that tan tan, I represent arcsin function
Asin, R represent that inverse cosine function acos, N represent that arctan function atan, P represent that power function pow, Q represent square root function
Sqrt, E represent that natural number is bottom exponential function exp, A represent that ABS function abs, G represent that sign function sign, L are represented
Natural logrithm function ln, B represent that denary logarithm function lg, D represent that hyperbolic sine function sinh, F represent hyperbolic cosine
Function cosh, H represent hyperbolic tangent function tanh；
I.e. functor collection F is：F=Sin, cos, tan, asin, acos, atan, pow, sqrt, exp, abs,
Sign, ln, lg, sinh, cosh, tanh }
Terminal symbol collection includes multidimensional terminating character and constant terminating character collection in this step；
Wherein multidimensional terminating character T is：
T={ a, b, c, d, e, f, g, h, i }；
Each element in wherein multidimensional terminal symbol collection T represents each independent variable in multidimensional variable respectively, in inspection human body fortune
By in the actual value substitution modelling of human body motion corresponding to each variable during movable model；Wherein constant terminal symbol collection C is：C=
{ 1000, -1000 }.
Loop termination condition is usually arranged as one of following condition in this step：
(1) the evolution number of times for representing the population of the whole human motion motion model of construction has reached certain value；
(2) adaptive value for representing the population optimum individual of whole human motion motion model reaches value set in advance, or
It is not changed in certain evolution number of times.
S22, the random artis motion sequence initial population of establishment, are gene expression by each artis motion sequence in population
Formula tree, i.e. ET trees；Wherein initial population is initial modelling of human body motion, is made up of each artis motion sequence, and joint
Point motion sequence is made up of functor collection and terminal symbol collection again；Wherein terminal symbol is concentrated includes artis three-dimensional time sequence number
According to middle time t, artis three-dimensional time sequence data coordinate components etc.；
The process that each artis motion sequence expression formula is changed into gene expression tree in this step is as follows：
Root node of the first symbol of the artis motion sequence that will be generated at random as expression formula；
By the character of artis motion sequence by the method for breadth traversal it is successively in order each element set functor
Or terminal symbol generation tree node, wherein, the number of child node is the number (terminal symbol and without ginseng function that the node answers containing parameter
Required number of parameters is 0, and the typical value of child node is according to symbol order filling polishing in sequence.
Artis motion sequence in the present embodiment represents the gene in evolution algorithmic, such as i-th artis is Ji, certain
At a time the coordinate data of the t artis is under individual behaviorBy i-th joint
The three-dimensional time sequence data of point can get, such asIt is one
Legal sequence, wherein Q representative functions accord with evolution, and the part with underscore is gene afterbody, and the part without underscore is base
Because of head.The corresponding expression trees of Fig. 3.Wherein Fig. 4 A to 4G show the generating process of the expression tree of Fig. 3.
S23, the adaptive value for calculating each artis motion sequence in population, then estimate model or absolute by relative error
Error model is estimated to each artis motion sequence in population to the adaptability of environment, qualified in assessment result
In the case of, a number of individuality is selected as male parent from population；And judge whether to meet loop termination condition, if full
Foot, then jump to S26, otherwise into step S24；
Relative error wherein used in this step estimates that model is：
Absolute error wherein used in this step estimates that model is：
Wherein n is the sum of the artis motion sequence checking sample of extraction in population, fiRepresent artis JiMotion sequence
Fitness value, R be a constant, (0, R) be fitness value control scope；P(i,j′)It is using artis JiCorresponding motion
Calculated by sequencej′Individual artis motion sequence verifies the functional value of sample, Yj′It isj′Individual artis motion sequence checking
The actual value of sample；
Using two norms (refer to spatially two air line distance of vector matrix), shape is such as:It is current to evaluate
The quality of model, whereinIt is predicted vector, Y is actual value vector.Work as f in this stepiWhen being substantially equal to n*R, i.e., relatively
Model of error estimate and absolute error are estimated in model | P(i,j′)-Yj′| it is substantially equal in the case of 0, then in this step
Judge that assessment result is eligible.
S24, the male parent selected according to step S23, the intersection of gene expression programming are carried out to the individuality in population, are become
ETTHER-OR operation；
S25, population at individual is carried out TIS insert string operation, obtain population at individual of future generation, return to step S23；
S26, output model, obtain the modelling of human body motion of the function expression represented by optimal chromosome, i.e. artis.
Artis motion sequence is made up of head and afterbody, and the length of artis motion sequence refers to the length of head and afterbody
Degree sum, effective length is generally less than the length of artis motion sequence in artis motion sequence.If artis motion sequence
The length of head is h, and the length of afterbody is I, then the length h of artis motion sequence head and artis motion sequence afterbody
Relation between length I can be defined as：
I=h (n-1)+1；
Wherein, in above formula, the parameter mesh number quantity of the most functor of institute's containing parameter mesh number in the symbol set of n representative functions.
So regulation is to ensure that the length of artis motion sequence is enough and sequence will not cause illegal showing due to length reason
As.
Wherein the calculation procedure of effective length is as follows in artis motion sequence：
1) first symbol that two pointers p, q point to artis motion sequence is set up；
If 2) what pointer p was pointed to is functor, moved after pointer q, the digit moved afterwards parameter for needed for the functor
Number；
3) 1 is moved after pointer p；
If 4) pointer p is not after pointer q, the 2) step is returned；
5) the effective length digit of artis motion sequence is calculated, finger is played from first sign bit of artis motion sequence
Sign bit pointed by pin q stops as ordered sequence, and its length is the effective length of sequence；
Above-mentioned TIS inserts string operation process including as follows in the present embodiment：
As artis motion sequence effective length m=1, i.e. the root position of artis motion sequence is a constant, this
When, it is the functor string for getting at random to be made a variation to [L/4] position since artis motion sequence root position；Wherein [L/4] table
Show L/4 round numbers part；Other characters of artis motion sequence head are moved afterwards successively simultaneously, beyond the character string of head length
Delete；Wherein L is the length of artis motion sequence, and h is the length of artis motion sequence head.
For example certain artis motion sequence is：ab*/aba+*aababbbbbaba, wherein band horizontal line part is transported for artis
The afterbody of dynamic sequence, and be not the head of artis motion sequence with horizontal line part, a, b, c represent artis J respectivelyiCoordinate
DataThe length L=21 of the artis motion sequence, its head length h=10, tail length I=11, lead to
Crossing the sequence can draw its effective length m=1, meet m=1, and string operation is inserted according to above-mentioned TIS, to the since root position
The variation of [L/4] position is the functor string that gets at random, will first position of artis motion sequence start to the 5th position
It is the function string being randomly derived to put variation, if the function string being randomly derived is * -+QS, carries out TIS and inserts string operation
Afterwards, the artis motion sequence is changed into * -+QS ab*/aababbbbbaba, wherein first position of artis motion sequence
Start to the 5th position to make a variation while for * _ * SQ, moved after the ab*/aba+* of its head, beyond head length after moving afterwards
Character string ba+* is deleted.By TIS insert string operation after artis motion sequence * -+QS ab*/aababbbbbaba, can be with
Learn that the significance bit of the pass artis motion sequence becomes for 13.
When artis motion sequence imitates length 1<m<During h, in last functor of artis motion sequence significance bit
The random functor string for getting of insertion below h-m, other characters behind artis motion sequence head insertion point are moved afterwards successively,
Character string beyond head length is deleted；
For example certain artis motion sequence is：+b/ababa+*ababbbbaba, wherein band horizontal line part is transported for artis
The afterbody of dynamic sequence, and be not the head of artis motion sequence, the length L=of the artis motion sequence with horizontal line part
21, its head length h=11, tail length I=10 can draw its effective length m=5 by the sequence, and significance bit is+b/
Ab, meets 1<m<H, string operation is inserted according to above-mentioned TIS, each random in last functor/rear insertion h-m=5 of significance bit
The function string for obtaining, if the function string being randomly derived is * _ * SQ, after carrying out the slotting string operations of TIS, artis fortune
Dynamic sequence is changed into+b/*_*SQ abababbbbaba, after wherein * _ * SQ insertions, moved after the ababa+* of head, exceed after moving afterwards
The character string aba+* of head length is deleted.It is+b/*_*SQ that the artis motion sequence after string operation is inserted by TIS
abababbbbaba, can learn that the significance bit of the artis motion sequence is changed into 14, effective length is more long more to can guarantee that kind
The uniformity of the individual distribution of group, can preferably avoid algorithm from sucking locally optimal solution.
When artis motion sequence effective length is m>=h, any operation is not done；
From above-mentioned, the present embodiment is used by gene expression programming of the prior art (GEP) intersecting, variation behaviour
TIS is added after work insert string operation construction and obtain self feed back gene expression programming sample is modeled, in modeling process
Inserting string operation by TIS increases the length of artis motion sequence significance bit, such that it is able to increase the uniformity of population at individual,
The globally optimal solution for making it more can quickly converge to problem, it is to avoid be absorbed in locally optimal solution.Therefore the present embodiment method is used
Self feed back gene expression programming is modeled the degree of accuracy that can effectively improve Human bodys' response and identification to sample
Speed.Meanwhile, the present embodiment gradient information representative for specific behavior is special as the model in modelling of human body motion
Levy, compared to existing technology the middle aspect of model extracted as the aspect of model, the present embodiment using global or local motion feature
Human body behavior campaign can more accurately be reflected, greatly reduce feature complexity, improved feature representativeness and identification is correct
Rate.
Above-described embodiment is the present invention preferably implementation method, but embodiments of the present invention are not by above-described embodiment
Limitation, it is other it is any without departing from Spirit Essence of the invention and the change, modification, replacement made under principle, combine, simplification,
Equivalent substitute mode is should be, is included within protection scope of the present invention.
Claims (10)
1. a kind of Human bodys' response method based on self feed back gene expression programming, it is characterised in that step is as follows：
S1, acquisition human body behavior depth image, then extract the N number of artis of human body each respectively from human body behavior depth image
Self-corresponding three-dimensional time sequence data, as a sample；
S2, sample is modeled using self feed back gene expression programming, gets the corresponding each artis motion sequence of sample
Row, so as to obtain the modelling of human body motion that each sample is based on artis；Wherein described self feed back gene expression programming is by gene
Expression formula is programmed in after intersection, mutation operation to add TIS and insert string operation construction and obtains, and wherein TIS inserts string operation and refers to
Artis motion sequence head inserts functor string, and the length for increasing artis motion sequence significance bit increases；
S3, the gradient information of extraction modelling of human body motion are used as the aspect of model；
S4, training sample set is got by step S1, and each training sample gets each training by way of step S2
The modelling of human body motion of sample, the model that the modelling of human body motion of each training sample is then extracted by way of step S3 is special
Levy；
S5, the aspect of model of the modelling of human body motion of each training sample are instructed as the input of neutral net to neutral net
Practice, the neural network model for obtaining will be trained, as final human body behavior grader；
S6, test sample is got by step S1, and test sample gets test sample by way of step S2
Modelling of human body motion, then extracts the aspect of model of the modelling of human body motion of test sample by way of step S3；
S7, the aspect of model of the modelling of human body motion of test sample is input into the human body behavior grader got into step S5
In, Human bodys' response result is obtained by the human body behavior grader.
2. the Human bodys' response method based on self feed back gene expression programming according to claim 1, its feature exists
In, in the step S1 use Microsoft's somatosensory device Kinect combination second generation SDK and OpenCV computer visions storehouse obtain
Human body behavior depth image；TOF (Time Of Flight, flight time) technology is wherein utilized, Microsoft Kinect sensings are calculated
The IR of infrared sensor transmitting reaches the phase difference of human body back reflection in device, obtains the human body behavior depth based on human body
Image；
The corresponding three-dimensional time sequence data of each artis is the infrared sensor with Microsoft somatosensory device Kinect as origin
What the rectangular coordinate system in space set up was obtained, the rectangular coordinate system in space includes x-axis, y-axis and z-axis, and wherein x-axis positive direction is
Parallel to Microsoft somatosensory device Kinect horizontal directions to the left, y-axis positive direction is to be inclined perpendicular to Microsoft somatosensory device Kinect
Upwards, z-axis positive direction is Microsoft's somatosensory device Kinect directions in direction.
3. the Human bodys' response method based on self feed back gene expression programming according to claim 1, its feature exists
In N is more than 25 in the step S1.
4. the Human bodys' response method based on self feed back gene expression programming according to claim 1, its feature exists
In in the step S1 after human body behavior depth image is got, the human body in human body behavior depth image being entered with background
Row segmentation, is then converted to skeleton tracing system, so as to obtain multiple artis three-dimensional time sequence datas of human body.
5. the Human bodys' response method based on self feed back gene expression programming according to claim 1, its feature exists
In self feed back gene expression programming is modeled to sample in the step S2, obtains the human motion mould based on artis
The detailed process of type is as follows：
S21, determine modelling of human body motion needed for functor collection and terminal symbol collection and terminal symbol number, and set loop termination
Condition；
S22, the random artis motion sequence initial population of establishment, are gene expression tree by the expression of each artis motion sequence；
Artis motion sequence is made up of functor collection and terminal symbol collection, and terminal symbol concentration is included in artis three-dimensional time sequence data
The coordinate components of time t and artis three-dimensional time sequence data；
S23, the adaptive value for calculating each artis motion sequence in population, then estimate model or absolute error by relative error
Model is estimated to each artis motion sequence in population to the adaptability of environment, in the qualified situation of assessment result
Under, a number of individuality is selected as male parent from population；And judge whether to meet loop termination condition, if meeting,
S26 is jumped to, otherwise into step S24；
S24, the male parent selected according to step S23, gene expression programming is carried out to the artis motion sequence in population
Intersection, mutation operation；
S25, the artis motion sequence to population carry out TIS and insert string operation, obtain population at individual of future generation, return to step S23；
S26, output model, obtain the modelling of human body motion of the function expression represented by optimal chromosome, i.e. artis.
6. the Human bodys' response method based on self feed back gene expression programming according to claim 5, its feature exists
In functor collection F is in step S21：
F={ S, C, T, I, R, N, P, Q, E, A, G, L, B, D, F, H }
It is as follows that each element in wherein functor collection F represents implication：
S represents that SIN function Sin, C represent that cosine function cos, T represent that tan tan, I represent arcsin function asin, R
Represent that inverse cosine function acos, N represent that arctan function atan, P represent that power function pow, Q represent square root function sqrt, E table
Show that exponential function exp, A that natural number is bottom represent that ABS function abs, G represent that sign function sign, L represent natural logrithm
Function ln, B represent that denary logarithm function lg, D represent that hyperbolic sine function sinh, F represent hyperbolic cosine function cosh,
H represents hyperbolic tangent function tanh；
Terminal symbol collection described in step S21 includes multidimensional terminating character and constant terminating character collection；
Wherein multidimensional terminating character T is：
T={ a, b, c, d, e, f, g, h, i }；
Each element in wherein multidimensional terminal symbol collection T represents each independent variable in multidimensional variable respectively, in inspection human motion mould
By in the actual value substitution modelling of human body motion corresponding to each variable during type；
Wherein constant terminal symbol collection C is：C={ 1000, -1000 }.
7. the Human bodys' response method based on self feed back gene expression programming according to claim 1 or 6, its feature
It is that the process that TIS inserts string operation is as follows：
As artis motion sequence effective length m=1, i.e. the root position of artis motion sequence is a constant, now, from
It is the functor string for getting at random that artis motion sequence root position starts to the variation of [L/4] position；Wherein [L/4] represents L/4
Round numbers part, L is the length of artis motion sequence；Other characters of artis motion sequence head are moved afterwards successively simultaneously,
Character string beyond head length is deleted；Wherein L is the length of artis motion sequence, and h is artis motion sequence head
Length；
When artis motion sequence imitates length 1<m<During h, behind last functor of artis motion sequence significance bit
The random functor string for getting of insertion h-m, other characters behind artis motion sequence head insertion point are moved, exceeded afterwards successively
The character string of head length is deleted.
8. the Human bodys' response method based on self feed back gene expression programming according to claim 1, its feature exists
In in the step S3, for each artis J of samplei(i=1,2..., N), N is the sum of artis in sample, is remembered
The modelling of human body motion for arriving is zi=Fi(x,y,t)；The aspect of model detailed process of modelling of human body motion is extracted in step S3 such as
Under：
Equidistantly taking k moment is designated as tj, j=1,2 ..., k；
Calculate the gradient information of each artisWithI=1,2 ..., N；J=1,2 ..., k；Wherein
Represent i-th artis in t respectivelyjMoment corresponding x, y-coordinate value；
By the above-mentioned each artis gradient information being calculatedWithI=1,2 ..., N；J=1,2 ..., k makees
For modelling of human body motion is zi=FiThe aspect of model of (x, y, t).
9. the Human bodys' response method based on self feed back gene expression programming according to claim 8, its feature exists
In the aspect of model of the modelling of human body motion in the step S5 by each training sample trains the tool for obtaining neural network model
Body process is as follows：
S51, neural networks with single hidden layer framework is initially set up, the mould of the modelling of human body motion of each training sample will be obtained in step S4
Type featureI=1,2 ..., N；J=1,2 ..., k as neutral net input training sample set, its
Middle to set N × 2k input neutral net, it is N × 4k+1 to set hidden neuron number；According to the class that human body behavior includes
Not Shuo M to set output neuron number be that M, i.e. output vector are { c1,c2,…,cs,…,cM, wherein cs∈ R, and value area
Between be [0,1], work as cs=max { c1,c2,…,cMWhen, cs1, i.e. neutral net output vector are taken as { 0,0 ..., 1 ..., 0 },
Represent that human body behavior belongs to s class behaviors；
S52, to set neural network learning rate η be the random number between 0~1, and neural network transformation function f (x) is set to
Sigmoid functions, learning rules are continuous perceptron learning rules, and neutral net is trained using error backpropagation algorithm, are obtained
To neural network model.
10. the Human bodys' response method based on self feed back gene expression programming according to claim 1, its feature exists
In, in the step S5, the aspect of model of the modelling of human body motion of each training sample is directed to, as the input of neutral net,
Cross-validation method training is torn open by ten and obtain neural network model, as final human body behavior grader.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710059525.6A CN106909891B (en) | 2017-01-24 | 2017-01-24 | A kind of Human bodys' response method based on self feed back gene expression programming |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710059525.6A CN106909891B (en) | 2017-01-24 | 2017-01-24 | A kind of Human bodys' response method based on self feed back gene expression programming |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106909891A true CN106909891A (en) | 2017-06-30 |
| CN106909891B CN106909891B (en) | 2019-11-15 |
Family
ID=59207616
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710059525.6A Active CN106909891B (en) | 2017-01-24 | 2017-01-24 | A kind of Human bodys' response method based on self feed back gene expression programming |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106909891B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107480692A (en) * | 2017-07-06 | 2017-12-15 | 浙江工业大学 | A kind of Human bodys' response method based on principal component analysis |
| CN108805965A (en) * | 2018-05-18 | 2018-11-13 | 合肥工业大学 | Human body physical motion generation method based on multi-target evolution |
| CN110325965A (en) * | 2018-01-25 | 2019-10-11 | 腾讯科技（深圳）有限公司 | Object processing method, equipment and storage medium in virtual scene |
| WO2021184792A1 (en) * | 2020-03-19 | 2021-09-23 | 中国科学院深圳先进技术研究院 | Dynamic balance assessment method and apparatus, and device and medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102764167A (en) * | 2012-06-12 | 2012-11-07 | 天津大学 | Myoelectric prosthesis control source lead optimization method based on correlation coefficients |
| CN104850846A (en) * | 2015-06-02 | 2015-08-19 | 深圳大学 | Human behavior recognition method and human behavior recognition system based on depth neural network |
| CN105615890A (en) * | 2015-12-24 | 2016-06-01 | 西安交通大学 | Angle and myoelectricity continuous decoding method for human body lower limb walking joint |
-
2017
- 2017-01-24 CN CN201710059525.6A patent/CN106909891B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102764167A (en) * | 2012-06-12 | 2012-11-07 | 天津大学 | Myoelectric prosthesis control source lead optimization method based on correlation coefficients |
| CN104850846A (en) * | 2015-06-02 | 2015-08-19 | 深圳大学 | Human behavior recognition method and human behavior recognition system based on depth neural network |
| CN105615890A (en) * | 2015-12-24 | 2016-06-01 | 西安交通大学 | Angle and myoelectricity continuous decoding method for human body lower limb walking joint |
Non-Patent Citations (2)
| Title |
|---|
| 李康顺等: "NCRS-GEP编程算法及其在统计建模中的应用", 《武汉理工大学学报》 * |
| 莫海芳等: "基因表达式编程的一种新遗传算子", 《计算机工程与应用》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107480692A (en) * | 2017-07-06 | 2017-12-15 | 浙江工业大学 | A kind of Human bodys' response method based on principal component analysis |
| CN110325965A (en) * | 2018-01-25 | 2019-10-11 | 腾讯科技（深圳）有限公司 | Object processing method, equipment and storage medium in virtual scene |
| CN108805965A (en) * | 2018-05-18 | 2018-11-13 | 合肥工业大学 | Human body physical motion generation method based on multi-target evolution |
| CN108805965B (en) * | 2018-05-18 | 2022-03-15 | 合肥工业大学 | Human physical motion generation method based on multi-objective evolution |
| WO2021184792A1 (en) * | 2020-03-19 | 2021-09-23 | 中国科学院深圳先进技术研究院 | Dynamic balance assessment method and apparatus, and device and medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106909891B (en) | 2019-11-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106845499A (en) | A kind of image object detection method semantic based on natural language | |
| CN106033435B (en) | Item identification method and device, indoor map generation method and device | |
| CN108334830A (en) | A kind of scene recognition method based on target semanteme and appearance of depth Fusion Features | |
| CN107423398A (en) | Exchange method, device, storage medium and computer equipment | |
| CN109271888A (en) | Personal identification method, device, electronic equipment based on gait | |
| CN109948526A (en) | Image processing method and device, detection device and storage medium | |
| CN107330444A (en) | A kind of image autotext mask method based on generation confrontation network | |
| CN106909891B (en) | A kind of Human bodys' response method based on self feed back gene expression programming | |
| CN110222580A (en) | A kind of manpower 3 d pose estimation method and device based on three-dimensional point cloud | |
| CN106650827A (en) | Human body posture estimation method and system based on structure guidance deep learning | |
| CN105243154B (en) | Remote sensing image retrieval method based on notable point feature and sparse own coding and system | |
| CN110246181B (en) | Anchor point-based attitude estimation model training method, attitude estimation method and system | |
| CN111222486B (en) | Training method, device and equipment for hand gesture recognition model and storage medium | |
| CN105051755A (en) | Part and state detection for gesture recognition | |
| CN109614899B (en) | Human body action recognition method based on lie group features and convolutional neural network | |
| CN109919245A (en) | Deep learning model training method and device, training equipment and storage medium | |
| CN107066979A (en) | A kind of human motion recognition method based on depth information and various dimensions convolutional neural networks | |
| CN109508686A (en) | A kind of Human bodys' response method based on the study of stratification proper subspace | |
| CN109711356A (en) | A kind of expression recognition method and system | |
| CN109086664A (en) | A kind of polymorphic gesture identification method of sound state fusion | |
| Wang et al. | Labanotation generation from motion capture data for protection of folk dance | |
| Chen et al. | Improved convolutional neural network based scene classification using long short-term memory and label relations | |
| Wang | [Retracted] Long Jump Action Recognition Based on Deep Convolutional Neural Network | |
| CN112507940B (en) | Bone action recognition method based on differential guidance representation learning network | |
| Shaikh et al. | Comprehensive Study on Emotion Detection with Facial Expression Images Using YOLO Models. |
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 |