CN102663374B - Multi-class Bagging gait recognition method based on multi-characteristic attribute - Google Patents

Abstract

The invention relates to a multi-class Bagging gait recognition method based on a multi-gait characteristic attribute, which belongs to the technical field of pattern recognition. According to the method, a nearest neighbor classifier is used as a weak classifier, and an integration classifier is constructed by expanding a two-class attribute Bagging method to a plurality of classes on the basis of 20 gait attribute characteristic sets after wavelet packets are decomposed and principal components are completely analyzed so as to carry out gait identity identification. The method comprises the following steps of: preprocessing, extracting characteristics and finally classifying test samples by using a method combining a nearest neighbor classifying principle and an MCAB algorithm. According to the multi-class Bagging gait recognition method based on the multi-gait characteristic attribute, a method fusing wavelet packet decomposition (WPD) and (2D) 2 principal component analysis (PCA) is adopted for the first time to extract and also select gait characteristics. The problem of loss of high-frequency components in a traditional gait recognition method based on wavelet transformation or overlarge dimensionality caused by simply adopting all data is solved. The multi-class Bagging gait recognition method based on the multi-gait characteristic attribute has higher recognition rate and visual angle change robustness.

Description

Based on the multi-class Bagging gait recognition method of many characteristic attributes

Technical field

The invention belongs to mode identification technology, be specifically related to a kind of multi-class Bagging Gait Recognition based on multistep state characteristic attribute, being a kind of automatic analysis that utilizes the gait that computer technology, Digital image processing technique, pattern-recognition etc. realize people and sentence method for distinguishing, is the algorithm about Method of Gait Feature Extraction and identification in living things feature recognition field.

Background technology

Biometrics identification technology refers to physiological characteristic that the mankind itself have, that can identify its identity or the technology that behavioural characteristic is carried out authentication utilized.Compared with traditional identity validation technology, biometrics identification technology has fundamentally been stopped forge and steal, and has higher reliability, security, is more and more widely used in the authentication of some security systems.

Gait Recognition technology is as a kind of novel biometrics identification technology, and it is the biological identification technology that the posture of walking according to people in video sequence is carried out identification.Compare with other biometrics identification technology, Gait Recognition technology is with its non-infringement, remote identity and the advantage such as be difficult to hide and be subject to the people's favor, has a wide range of applications in fields such as national public safety, financial security, authentication, video monitorings.

About the Feature Extraction Technology of gait, there is document to adopt WAVELET PACKET DECOMPOSITION to solve preferably this problem, but the characteristics of image dimension after WAVELET PACKET DECOMPOSITION is higher, and it adopts classical PCA algorithm to carrying out feature extraction, while adopting the method for svd to ask the eigenwert of correlation matrix and proper vector, calculate expend large.Two-dimensional principal component analysis (2DPCA) can directly calculate image data matrix, calculated amount is relatively few a lot, but after 2DPCA, need n*k (wherein, n is image resolution ratio, k is the rear selected characteristic column vector number of conversion, and k < n) individual data carrys out presentation video, and the dimension of proper vector is still higher.Principal component analysis (PCA) ((2D) completely ²pCA) can further reduce the dimension of proper vector, expend thereby reduce identification, and suitable with 2DPCA on its recognition performance, be even better than 2DPCA.

Someone proposes a kind of attribute Bagging algorithm AB about two class problems at present, the present invention is on the basis of the attribute Bagging of two class problems algorithm AB, proposes a kind of multi-class Bagging Gait Recognition based on multistep state characteristic attribute---MCAB (Multi-class Attribute Bagging) algorithm.

Summary of the invention

Content of the present invention is to have proposed multi-class Bagging (MCAB) Gait Recognition based on multistep state characteristic attribute.The method uses 1NN as Weak Classifier, by two generic attribute Bagging methods being expanded to the multi-class integrated classifier MCAB (Multi-class Attribute Bagging) that builds.We evaluate and test the method on NLPR gait data storehouse, and result shows, with simple employing wavelet packet and (2D) ²the recognition methodss such as PCA are compared, and this method has higher discrimination and visual angle change robustness.

Technology contents of the present invention is as follows:

For utilizing MCAB algorithm to test, we need, first by pretreated normalization gait image sequence is carried out to cycle detection, extract gait energygram, to overcome gait data amount problems of too; Again gait energygram is carried out to WAVELET PACKET DECOMPOSITION and principal component analysis (PCA) completely, the result images obtaining has represented respectively the not feature of ipsilateral of gait image; The last classification performance according to aforementioned different characteristic, also classifies with MCAB algorithm by the different attribute that each feature is considered as to gait.

Multi-class Bagging (MCAB) Gait Recognition based on multistep state characteristic attribute, the step of the method comprises: the pre-service of body gait sequence, feature extraction, finally utilize MCAB algorithm that test sample is grouped in corresponding class, and recognition effect is evaluated, its concrete steps are as follows:

Step 1, pre-service

(1) morphology processing

The humanbody moving object image of background separation is carried out to morphology processing, and the cavity existing to remove binary image, obtains more excellent segmentation effect;

(2) target is extracted

Utilize the method for 8 connected component analyses to extract a simply connected moving target, people's silhouette is removed residual noise, thereby obtains more excellent two-value profile diagram;

(3) image normalization

Cut out the gait image of standard according to human body contour outline coordinate, obtain size normalized image, wherein, the size of image is unified is 64*64 pixel.

Step 2, feature extraction

(1) detection of gait cycle

Utilize the profile width of human body that the characteristic sexually revising synchronizing cycle occurs in time, carry out dividing gait cycles by the wide variety signal of human body contour outline;

(2) set up gait energygram (GEI), the representative using gait energygram as different gait sequences;

Gait energygram: after carrying out gait cycle detection, by the gait sequence image processing generation GEI in one-period be:

$G(x,y)=\frac{1}{N_1}\underset{t=1}{\overset{N_1}{\mathrm{\&Sigma;}}}{B}_t(x,y)$

In formula, G (x, y) is the gait energygram of this periodic sequence, N ₁the length of complete gait cycle sequence, B _t(x, y) is t gait image in one-period, x, and y represents two dimensional image plane coordinate.

(3) merge WPD+ (2D) ²pCA feature selecting

First, adopt secondary WAVELET PACKET DECOMPOSITION (WPD) gait energygram, after decomposing, every width gait energygram obtains 20 images after decomposition; Then, adopt respectively complete principal component analysis (PCA) ((2D) ²pCA) (two-dimensional principal component analysis of column direction at once) further extracts validity feature; Finally identify respectively with nearest neighbor classifier, respectively each feature is carried out the calculating of discrimination.

Step 3, Ensemble classifier identification

(1) determine the primitive attribute set of double sampling

In MCAB algorithm, first need the community set of original training example to carry out there is the double sampling of putting back to for n time, n is process WPD+ (2D) ²after PCA, discrimination is more than or equal to 50% attribute number, forms new training example by this n attribute, then by these new training instance constructs Weak Classifiers.Owing to primitive attribute collection being had to the double sampling of putting back to, the attribute therefore having in new training example is described may occur repeatedly, and some attributes may once also not occur.We select those discriminations to be more than or equal to 50% attribute formation community set AttributeSet.

(2) utilize MCAB algorithm to carry out Classification and Identification

Have document to propose a kind of attribute Bagging algorithm AB about two class problems, be provided with two classifications, classification space is Y={1 ,-1}, and instance space is ￠, training example set is S={ (x ₁, y ₁), (x ₂, y ₂) ..., (x _m, y _m), wherein x _i∈ ￠, y _i∈ Y, example x _iwith n attribute representation, property set A is { a ₁, a ₂..., a _n.Algorithm AB carries out T wheel altogether, and each is concentrated and have a double sampling n attribute of putting back to from primitive attribute in taking turns, and forms new training example collection, then will newly train example base classification algorithm training to go out base sorter h by the training example of this n new attribute description _t(x) → { 1,1} distributes to h _t(x) weights are a _t; Finally unknown example is voted by the result of T base sorter, simple vote function is

H(x)＝sign(f(x)) (1)

Wherein, $f\left(x\right)=\underset{t=1}{\overset{T}{\mathrm{\&Sigma;}}}{a}_t{h}_t\left(x\right),$ a _t＝1。

But Gait Recognition problem is a multi-class classification problem, for this reason, we introduce M1 algorithm above-mentioned algorithm are expanded to the multi-class attribute Bagging of multistep state algorithm MCAB.If multi-class space is Y={1,2 ..., m}, y _i∈ Y.For multi-class M1 algorithm, establishing every training set of taking turns training use is S _t, wherein t=1 ..., T, wherein T is iterations, at S _tthe weak learning algorithm training of upper use generates Weak Classifier---anticipation function h _t(x) → Y, distributes to h _t(x) weights are a _t, last strong anticipation function is

$H\left(x\right)=\arg \underset{y\&Element;Y}{\max }\underset{t:h_t\left(x\right)=y}{\mathrm{\&Sigma;}}{a}_t\left|\right|h_t\left(x\right)=y\left|\right|---\left(2\right)$

Wherein, when the category label assigned to when base sorter is identical with y, || h _t(x)=y|| is 1, otherwise is 0;

r _tfor h _t(x) the correct number of classifying deducts the number of point classification error, and t is iterations, t=1 ..., T.

Above formula can be understood as test sample book x can assign to a classification y ∈ Y on each base sorter, and establishing y has weight w, if base sorter h _tassigned to y upper, corresponding weight w is this base sorter h _tweights a _t, that class of final weights sum maximum is the class that test sample book is assigned to.

MCAB carries out T wheel, generally get 10=< T <=100, each is taken turns a resampling n attribute of putting back to from aforesaid primitive attribute set A ttributeSet, forms new training example collection by the training example of this n new attribute description; On new training example collection, go out different base sorter h with 1NN Algorithm for Training _t(x); Finally form integrated classifier by above-mentioned multiple base sorters by formula (2).

Beneficial effect of the present invention is: 1. by gait cycle is calculated to gait energygram, this width gait energygram has comprised the gait information such as the profile of each gait image, frequency, phase place in gait cycle, when can guaranteeing not abandon gait feature, reduce pending gait image data volume, consume to reduce to calculate; 2. utilize based on WPD+ (2D) ²pCA method is extracted feature, solve the existing gait recognition method medium-high frequency component based on wavelet transformation and lose or simply adopt dimension problems of too due to total data, the method more exactly abstraction reaction movement human walking feature effective information and reduce the intrinsic dimensionality for identifying gait; 3. utilize MCAB algorithm to carry out Classification and Identification to improve Gait Recognition accuracy, and visual angle change is had to good robustness.

Accompanying drawing explanation

Fig. 1 is the process flow diagram of algorithm of the present invention.

Fig. 2 is pretreated process flow diagram in algorithm of the present invention

Fig. 3 is the process flow diagram that algorithm characteristics of the present invention is extracted

Fig. 4 is secondary WPD+ (2D) in algorithm of the present invention ²pCA feature extraction process flow diagram

Fig. 5 is wavelet packet exploded view in algorithm of the present invention

Embodiment

Provide in detail the explanation of each detailed problem related in this invention technical scheme below:

Step 1, preprocessing process are as follows:

The database that we adopt is a gait data storehouse of Institute of Automation Research of CAS, and this database is by background separation, and the work that the present invention will do is on this basis, to carry out pre-service, thereby carries out cycle detection and calculate the operations such as gait energygram.

(1) morphology processing

Due to the impact of other extraneous factors such as weather, illumination, shadow, in the image after background separation, can there is unavoidably noise, therefore also need image to do further processing, to obtain best segmentation effect.The present invention eliminates the noise in bianry image with morphologic filtering and fills up the disappearance of moving target.As a kind of conventional image filter method for de-noising, morphology is dilation and corrosion for the fundamental operation of image filtering, mutually combining and deriving other two kinds of computings by dilation and corrosion: opening operation and closed operation.Opening operation can smooth object cam contour, disconnect narrow connection, remove tiny jut; Closed operation can smooth object concave contour, long and narrow breach is connected into tiny curved mouth.Utilize this character can realize the object of filtering and filling cavity.

(2) target is extracted

After morphology is processed, still may exist part clutter noise to form piece not of uniform size, and real moving target is maximum in these pieces often.Therefore image is further carried out to connected domain analysis, utilize the method for 8 connected component analyses to extract a simply connected moving target, object is only to retain the moving target in image, thereby obtains more excellent two-value profile diagram.

(3) image normalization

For the impact of removal of images size on identification, should first make human body placed in the middle, be then 64*64 pixel by the size unification of image.

Step 2, characteristic extraction procedure are as follows:

(1) detection of gait cycle

People's walking is the behavior of one-period, definition gait cycle for: again land the experienced time to corresponding leg heel from heel contact, comprise two stance phases and two shaking peroids.In order to raise the efficiency, the present invention utilizes the profile width of human body that the characteristic sexually revising synchronizing cycle occurs in time, carrys out dividing gait cycles by the wide variety signal of human body contour outline.

(2) gait energygram

After carrying out the cycle detection of gait, directly adopting gait cycle to carry out Gait Recognition exists data volume huge, aggravate the difficulty of Method of Gait Feature Extraction and calculated to consume, for guarantee that not abandoning gait feature reduces pending gait image data volume simultaneously, the present invention has adopted the method for gait energygram, the some gait images that are about in a cycle synthesize piece image through average weighted method, and this width image has comprised in gait cycle the gait information such as the profile of each gait image, frequency, phase place.This method does not need the size at the gait interval of considering every frame, and avoids the impact of some accidentalia.For given two-value gait cycle image sequence B _t(x, y), GEI is defined as follows:

$G(x,y)=\frac{1}{N_1}\underset{t=1}{\overset{N_1}{\mathrm{\&Sigma;}}}{B}_t(x,y)$

Wherein, G (x, y) is the gait energygram of this periodic sequence, N ₁the length of complete gait cycle sequence, B _t(x, y) is t gait image in one-period, x, and y represents two dimensional image plane coordinate.

(3) merge WPD+ (2D) ²pCA feature selecting

Every two field picture is divided into low frequency component A and high fdrequency component by two-dimensional discrete wavelet conversion, and comprising horizontal component H, vertical component V and diagonal components D, the low frequency component A that the method has only retained picture signal, has lost high fdrequency component.WAVELET PACKET DECOMPOSITION (WPD) has not only retained the low frequency component of picture signal, has also retained high fdrequency component.

Characteristics of image dimension after WAVELET PACKET DECOMPOSITION is higher, adopts classical PCA algorithm when carrying out feature extraction, and while adopting the method for svd to ask the eigenwert of correlation matrix and proper vector, calculated amount is very large.And two-dimensional principal component analysis (2DPCA) directly calculates matrix, calculated amount is relatively few a lot.But 2DPCA also has its drawback aspect image characteristics extraction.Suppose that image size is 64*64, (wherein k is the rear selected characteristic column vector number of conversion after 2DPCA, to need 64*k, and k < 64) individual data carry out presentation video, in order to meet accuracy requirement, conventionally k is greater than 5, so at least need 320 data to carry out presentation video, the dimension of proper vector is still higher.Principal component analysis (PCA) ((2D) completely ²pCA) can further reduce the dimension of proper vector, expend thereby reduce identification, and suitable with 2DPCA on its recognition performance, be even better than 2DPCA.For the information of more effectively utilizing image is carried out Gait Recognition, based on above-mentioned consideration, a kind of fusion WAVELET PACKET DECOMPOSITION (WPD) and (2D) is proposed herein ²the method of PCA.

Specific implementation process is as follows:

First every width gait energygram G (x, y) is carried out to secondary WPD decomposition, obtain 20 images after decomposition.That is: utilize wavelet packet character to carry out first order WPD decomposition to original gait energygram, obtain the image of low frequency A1, horizontal high frequency H1, vertical high frequency V1, diagonal angle high frequency D1, respectively these four images are carried out to second level WPD decomposition again, obtain successively low frequency A2, horizontal high frequency H2, vertical high frequency V2, diagonal angle high frequency D2, low frequency HA2, horizontal high frequency HH2, vertical high frequency HV2, diagonal angle high frequency HD2, low frequency VA2, horizontal high frequency VH2, vertical high frequency VV2, diagonal angle high frequency VD2, low frequency DA2, horizontal high frequency DH2, vertical high frequency DV2, diagonal angle high frequency DD2.

Secondly, image after respectively WPD being decomposed (A1 ..., DD2) and carry out (2D) ²pCA conversion, (2D) ²the detailed process of PCA conversion is as follows:

I) calculated population sample average matrix

$\stackrel{\&OverBar;}{A}=\frac{1}{M}\underset{k=1}{\overset{M}{\mathrm{\&Sigma;}}}{A}_k$

Wherein, A _kbe k sample image, M is total sample number

Ii) calculate sample covariance matrix

$G_1=\frac{1}{M}\underset{k=1}{\overset{M}{\mathrm{\&Sigma;}}}(A_k-\stackrel{\&OverBar;}{A}){(A_k-\stackrel{\&OverBar;}{A})}^T$

$G_2=\frac{1}{M}\underset{k=1}{\overset{M}{\mathrm{\&Sigma;}}}{(A_k-\stackrel{\&OverBar;}{A})}^T(A_k-\stackrel{\&OverBar;}{A})$

Calculate respectively eigenwert and the standard orthogonal characteristic vector of covariance matrix G1, G2, the matrix of the proper vector composition of front d1, d2 larger nonzero eigenvalue is U, V.Wherein, $U={[u_1,u_2,\&CenterDot;\&CenterDot;\&CenterDot;,u_{d_1}]}^T,V=[v_1,v_2,\&CenterDot;\&CenterDot;\&CenterDot;,v_{d_2}]$

Iii) generate (2D) ²eigenmatrix Y ' after PCA conversion

Y′＝UAV

Wherein, A is sample matrix arbitrarily.

Finally, respectively 20 image sequences after conversion are carried out to arest neighbors classification, utilize 1NN to calculate discrimination to each image sequence respectively.

Step 3, utilize MCAB algorithm to classify

(1) determine the primitive attribute set of double sampling

Take 0 ° as example, in selection, in step, last discrimination discrimination is more than or equal to 50% feature and forms the primitive character set A ttribute Set={A1 that treats double sampling, H1, V1, A2, H2, V2, HA2, HH2}.Be provided with 20 people's gait image, classification number is 20.

(2) utilize MCAB algorithm to carry out Classification and Identification

Step1: determine that the number in double sampling community set is 8, determine iterations 10;

Step2:For t=1: 10, iteration 10 times;

Iteration for the first time: put back to and sample 8 times from Attribute Set, extract one at every turn, obtain property set A1={A1, H1, HH2, V2, H2, V1, H1, A1}, owing to being to have randomly the extraction of putting back at every turn, therefore some attribute may occur repeatedly, for example: A1, H1, some attribute may there will not be, for example: A2, HA2.Each training sample instance X _kattribute only get A ₁obtain vectorial S _1k=[A1 ^(k)h1 ^(k)hH2 ^(k)v2 ^(k)h2 ^(k)v1 ^(k)h1 ^(k)a1 ^(k)], X _kuse S _1krepresent, go out base sorter h with 1NN Algorithm for Training ₁(x) → Y, calculates weights:

$a_1=\frac{1}{2}\ln \left|\frac{1+r_1}{1-r_1}\right|$

Wherein, r ₁=h ₁(x) number of correct classification deducts h ₁(x) number of mis-classification.

The tenth iteration: put back to ground double sampling and obtain property set A 8 times from Attribute Set ₁₀={ V2, HH1, A2, H2, HH2, HA2, H1, A1}, each example sample X _kattribute only get A ₁₀obtain vectorial S _10k=[V2 ^(k)hH1 ^(k)a2 ^(k)h2 ^(k)hH2 ^(k)hA2 ^(k)h1 ^(k)a1 ^(k)]; X _iuse S _10krepresent, go out base sorter h with 1NN Algorithm for Training ₁₀(x) → Y, calculates weights

$a_{10}=\frac{1}{2}\ln \left|\frac{1+r_{10}}{1-r_{10}}\right|$

Wherein, r ₁₀=h ₁₀(x) number of correct classification deducts h ₁₀(x) number of mis-classification.

Step3: for any test sample book x,

$H\left(x\right)=\arg \underset{y\&Element;Y}{\max }\underset{t:h\left(x\right)=y}{\mathrm{\&Sigma;}}{a}_t\left|\right|h_t\left(x\right)=y\left|\right|$

Wherein, Y is classification set, y ∈ Y; T is iterations, and value is 1...T.

In Step3, formula further illustrates as follows: for any test sample book x, the classification situation of each base sorter is as follows:

h ₁(x)-＞1，h ₂(x)-＞2，h ₃(x)-＞2，h ₄(x)-＞3，h ₅(x)-＞1，h ₆(x)-＞2，h ₇(x)-＞5，h ₈(x)-＞19，h ₉(x)-＞1，h ₁₀(x)-＞3。

The weights sum of classification 1 is: a ₁+ a ₅+ a ₉;

The weights sum of classification 2 is: a ₂+ a ₃+ a ₆;

The weights sum of classification 3 is: a ₄+ a ₁₀;

The weights sum of classification 5 is: a ₇;

The weights sum of classification 19 is: a ₈;

All the other category label weights are: 0.

The category label of final test sample is: the maximum corresponding category label of weights sum,

That is: max{a ₁+ a ₅+ a ₉, a ₂+ a ₃+ a ₆, a ₄+ a ₁₀, a ₇, a ₈corresponding category label.When there being the weights sum of two or more maximums, give any one category label in these several kinds to test sample book.

Detailed description experimental result of the present invention below:

The database that experiment of the present invention adopts is the CASIA Dataset A database that Institute of Automation, CAS provides, this database comprises 20 people altogether, everyone has respectively 3 different visual angles (0 °, 45 ° and 90 °), 4 sequences are taken respectively at each visual angle, amount to and comprise 240 sequences.These color image sequences are taken with 25 frames speed per second, and original size is 352*240 pixel, and average length is about 100 frames.

In order to verify MCAB algorithm, we select arbitrarily three sequences as training data in NLPR storehouse in 4 sequences of everyone same view angle, and a remaining sequence is tested.

Table 1 is 20 people under three kinds of different visual angles (0 °, 45 °, 90 °) condition, adopts the basis function of Haar small echo as WAVELET PACKET DECOMPOSITION, and every number of sub images of WAVELET PACKET DECOMPOSITION is carried out respectively to (2D) ²pCA conversion, the result that finally adopts the aforesaid nearest neighbor classifier based on L2 normal form to identify, finally we have added up the asynchronous discrimination of iterations.

Table 1WPD+ (2D) ²the discrimination of PCA

Table 2 is take the recognition result of table 1 as basis, then calls the result that MCAB algorithm is identified, and finally we have added up the asynchronous discrimination of iterations.

The discrimination of table 2MCAB algorithm

Can obtain higher discrimination as can be seen from Table 2: the each subimage coefficient than simple use wavelet decomposition is identified, and in the time that iterations T is larger, the recognition performance of MCAB algorithm is higher, and to different visual angles (0 °, 45 °, 90 °).

Claims (2)

1. the multi-class Bagging gait recognition method based on multistep state characteristic attribute, comprise pre-service, the feature extraction of body gait sequence, last is that MCAB algorithm combines according to arest neighbors principle of classification with the multi-class Bagging based on multistep state characteristic attribute, test sample is grouped in corresponding class, it is characterized in that: concrete steps are as follows:

Step 1, pre-service

To the humanbody moving object image of background separation carries out morphology processing successively, target is extracted and image normalization processing;

Step 2, feature extraction

Through the detection of gait cycle, set up gait energygram, the representative using gait energygram as different gait sequences, utilizes WPD+ (2D) ²pCA method is calculated the discrimination of 20 features; WPD is secondary WAVELET PACKET DECOMPOSITION;

Step 3, Classification and Identification

Select discrimination in step 2 to be more than or equal to 50% feature as the primitive character set A ttribute Set that treats double sampling, utilize MCAB algorithm finally to identify test sample book; Specific as follows

(1) in MCAB algorithm, first need the community set of original training example to carry out there is the double sampling of putting back to for n time, n is that discrimination is more than or equal to 50% attribute number after step 2, this n attribute may occur repeatedly, what have may once also there will not be, form new training example by this n attribute, then by these new training instance constructs Weak Classifiers;

(2) utilize MCAB algorithm to carry out Classification and Identification

I: determine the number n of double sampling attribute, determine iterations T; Wherein n is the number that discrimination is more than or equal to 50% attribute, and T value is 10 ~ 100;

Ii:For t=1:T carries out following 3 steps;

A): from Attribute Set, a double sampling n attribute obtains property set A _t, to the each sample in training set S, only get A _tin attribute, obtain new attribute S _t;

B): at attribute S _tupper use 1 NN Algorithm for Training goes out base sorter h _t(x) → Y;

C): calculate weights

Wherein, r _t=h _t(x) number of correct classification deducts h _t(x) number of mis-classification;

Iii: for any test sample book x, integrated classifier H (x) is:

Wherein, Y is classification set, y ∈ Y; T is iterations, and value is 1 ... T.

2. the multi-class Bagging gait recognition method based on multistep state characteristic attribute according to claim 1, is characterized in that, the extraction step of described gait energygram is as follows:

After carrying out gait cycle detection, the gait energygram of processing generation by the gait sequence image in one-period is:

In formula, G (x, y) is the gait energygram of this periodic sequence, N ₁the length of complete gait cycle sequence, B _t(x, y) is t gait image in one-period, x, and y represents two dimensional image plane coordinate.

Images