CN104537340A - Novel gait cycle generating method - Google Patents

Abstract

The invention discloses a novel gait cycle generating method. The method comprises the steps of detecting all original gait cycles from a gait sequence to be processed according to periodic variation of pace width; setting a value k smaller than the minimum value among the numbers of frames in the original gait cycles to serve as the number of frames contained in a final cycle, wherein k belongs to multiples of 4; extracting k frames from each original gait cycle at equal intervals to serve as the representative frames of the cycle; obtaining the average frame of the representative frames on the same positions of the original gait cycles; calculating the coefficient of association between each representative frame in each gait cycle and the corresponding average frame, and selecting the representative frame with the highest coefficient of association to serve as the optimal frame of the corresponding position; obtaining the final gait cycle by means of the optimal frames on all positions. According to the method, incomplete frames and repeated frames in the gait sequence are eliminated, a gait cycle is generated by means of frames high in image quality, and then more accurate gait features are obtained and gait detection effect is effectively improved.

Description

A kind of new gait cycle generation method

Technical field

The present invention relates to a kind of Gait Recognition technology, particularly a kind of new gait cycle generation method.

Background technology

Gait Recognition is a kind of biometrics identification technology that the posture of being walked by people realizes differentiating individual identity.Its basic process is: first from original gait video, detect moving target, then calculates and generates gait cycle, then carry out feature extraction to the data representing gait cycle, finally feature is carried out Classification and Identification.Wherein, it is extremely important that gait cycle generates this link, whether accurately, whether representative is directly connected to acquired gait cycle data.Generate gait cycle accurately and effectively to have great significance to follow-up feature extraction and classifying identification.

The original gait image of one frame cannot comprise enough gait information, although one section of complete gait sequence comprising several gait cycles almost includes whole gait information, if the section of rounding gait sequence carries out calculating, time overhead can be caused to increase due to the existence of bulk redundancy information.Therefore Gait Recognition generally adopts a gait cycle to analyze.Obviously, generate gait cycle expression accurately to have great significance to Method of Gait Feature Extraction and Classification and Identification.

At present, domestic and international researcher has done large quantifier elimination in gait cycle generation method.The people such as BenAbdelkader published thesis on IEEE International Conference on AutomaticFace and Gesture Recognition in 2002 " Stride and Cadence as aBiometric in Automatic Person Identification and Verification ", proposed to utilize the bounding rectangles width of frame of human body contour outline to change and detected gait cycle.The major defect of the method is that the width of adopted bounding rectangles frame is always not consistent with the width of paces, causes the cyclical variation of the cyclical variation of border width and paces sometimes to there is certain deviation.The people such as Sarkar publish thesis " TheHumanID Gait Challenge Problem:Data Sets; Performance; and Analysis " at IEEETRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, to utilize in human walking procedure two lower limb variable area given rule to detect gait cycle; The people such as Wang Kejun use the change of border the ratio of width to height to detect gait cycle in the paper that 2009 deliver at Journal of Image and Graphics " ties up the gait recognition method of principal component analysis (PCA)s " based on gait energygram picture and 2.

The something in common of said method is that all frames of a use gait cycle are expressed as final gait cycle.But affect by factors such as illumination variation, object interference, background complicacy in reality, the moving target that common method for testing motion is difficult to ensure to detect all meets the requirements at all frames of each gait cycle, namely all likely occurs dissatisfactory frame (as incompleteness etc.) at any gait cycle.In addition, people, walk in process at any time may pause or slow down, causes comprising a large amount of repeating frames in gait sequence.If use these incomplete frames or repeating frame to express as the cycle, will make sample set can not completely true reflected sample attribute, affect recognition effect.

Summary of the invention

The object of the invention is to overcome the shortcoming of prior art and deficiency, a kind of new gait cycle generation method is provided, this gait cycle generation method can reject incomplete frame in gait sequence and repeating frame, the good frame of picture quality is utilized to generate a gait cycle, and then obtain more accurate gait feature, be conducive to the result promoting gait detection.

Object of the present invention is achieved through the following technical solutions: a kind of new gait cycle generation method, generate a gait cycle by choosing the best frame of picture quality in each gait cycle of same gait sequence, described gait cycle generation method comprises the steps:

Step 1: detect each original gait cycle from pending gait sequence;

Step 2: set the frame number k comprised in the expression of final cycle, this frame number should be less than the minimum value comprising frame number in each original gait cycle of step 1 gained;

Step 3: the frame number set according to step 2, equally spaced extracts the representative frame of k frame as this cycle from each original gait cycle;

Step 4: the quality assessment standard of definition gait two field picture, and its quality coefficient is calculated to all representative frame obtained by step 3;

Step 5: for the representative frame in same position in all gait cycles, chooses the optimal frames of the best frame of picture quality as this position, and the optimal frames of each position forms final gait cycle and expresses.

Can detect each original gait cycle according to the cyclical variation of paces width in described step 1, the frame comprised between adjacent two paces width maximum value or minimum values is half period, and continuous two semiperiods form a gait cycle.

The minimum value that the numerical value of setting in described step 2 should comprise frame number than each the original gait cycle calculated in step 1 is little, and this numerical value must be the integral multiple of 4.

Using 1/4th gait cycles as elementary cell in described step 3, each elementary cell is divided at equal intervals with set numerical value in step 2 1/4th, choose from the nearest frame of Along ent as the representative frame of place elementary cell on this position, namely the representative frame of adjacent four elementary cells forms the expression of a complete gait cycle.

The quality coefficient characterizing gait frame image quality quality in described step 4 is the related coefficient of the average frame of this frame and its position, and related coefficient is larger, then this frame quality is better; Average frame is by acquisition of averaging to the representative frame being in same position in each cycle.

According to the frame image quality judgment criteria that step 4 defines in described step 5, in comparison gait sequence, all gait cycles are in the quality level of the frame of correspondence position, choose the optimal frames of the best frame of wherein picture quality as this position, the optimal frames on all correspondence positions forms the final gait cycle expression that this gait sequence generates.

Position in described step 5 refers to the sequence number of certain representative frame in a current period k representative frame.

The present invention has following advantage and effect relative to prior art:

The present invention is directed to the gait cycle Generating Problems of the gait sequence comprising incomplete frame and repeating frame, overcome the shortcoming and defect of prior art, the information making full use of all gait cycles in gait sequence generates gait cycle more accurately and expresses.This invention removes the impact of the frame not meeting quality requirements in each gait cycle, maximally utilised more satisfactory frame in each cycle, thus make the final gait cycle generated express and can reflect actual conditions more exactly.The present invention not only can eliminate the impact brought by situations such as moving target speed change or pauses, also can reduce data redundancy and calculated amount, improves accuracy, is conducive to the result promoting gait detection.

Accompanying drawing explanation

Fig. 1 is FB(flow block) of the present invention.

Fig. 2 is one section of original gait sequence comprising repeating frame and defective frame.

Fig. 3 is 3 primitive periods detected from original gait sequence.

Fig. 4 represents in 3 cycles equally spaced selected from 3 primitive periods according to the frame number of setting.

Fig. 5 is the average frame in cycle representative frame on each correspondence position.

Fig. 6 is the gait cycle choosing optimal frames generation from all cycle representatives.

Embodiment

Below in conjunction with embodiment and accompanying drawing, the present invention is described in further detail, but embodiments of the present invention are not limited thereto.

Embodiment

As shown in Figure 1, as preferred embodiment, implementation process of the present invention is described using one section of gait sequence comprising N frame in CASIA (The Institute of Automation of the Chinese Academy ofSciences) DatasetB gait data storehouse.As shown in Figure 2, selected gait sequence comprises repeating frame (the 7 to 9 frame, the 35 to 37 frame etc.) and incomplete frame (the 1 to 4 frame, 49 to 55 frames and the 77 to 82 frame etc.).

Step 1: utilize the cyclical variation of paces width to detect each original gait cycle from gait sequence.Particularly, for n-th two-value gait frame, the horizontal ordinate x of the most left pixel of target area leg is found out _n,lwith the horizontal ordinate x of the rightest pixel _n,r, then this frame paces width is d _n=| x _n,l-x _n,r|.Same process is done to all frames of whole gait sequence, obtains paces width vector set d={d ₁, d ₂..., d _n.Analyze the extreme value of paces width vector set: to the differentiate of paces width vector set, detect its zero crossing.If the derivative adjacent with the zero crossing left side is greater than zero, be then maximum point, otherwise be minimum point.Frame between two adjacent maximum value or minimum values is half gait cycle, and two continuous semiperiods form a gait cycle.Such as, the frame between the 1st and the 3rd minimal value is the frame between primitive period the one, 3rd and the 5th minimal value is then the primitive period two, the like.As shown in Figure 3, be 3 primitive periods detected from original gait sequence.

Step 2: setting is used for the representative frame number that gait cycle is expressed.The frame number that each the original gait cycle obtained from step 1 comprises is also inconsistent, therefore needs setting fixing frame number.Set fixing frame number should be no more than the minimum value of each original gait cycle frame number of step 1 gained, and must be the integral multiple of 4.Because the frame number of primitive period 1,2,3 is respectively 28,28,26 in this example, so fixing frame number should be less than 26, and be the integral multiple of 4, be therefore set as 24 frames.

Step 3: the representative frame extracting fixed qty from each original gait cycle.Using 1/4th gait cycles (frame between a pair namely adjacent paces width maximum value and minimal value) as elementary cell.By 1/4th (in this example for 24/4=6) each elementary cell of decile setting frame number in step 2, choose from the nearest frame of Along ent as the representative frame of current basic unit on this position.The cycle that can obtain unified frame number thus expresses, and can remove the repeating frame in original gait cycle simultaneously.As shown in Figure 4, according to the frame number 24 of setting, equally spaced select 3 the cycle representatives comprising 24 frames from 3 primitive periods;

Step 4: the average frame calculating each position in the gait cycle that obtained by step 3.Suppose to comprise m gait cycle in gait sequence, according to the setting of step 2 in this example, there are 24 representative frame in each cycle.If f _i,jbe the jth representative frame in i-th cycle, then the average frame of each position is calculated as follows:

$\begin{array}{cc}{\stackrel{\‾}{f}}_j=\frac{1}{m}\underset{i=1}{\overset{m}{\mathrm{\Σ}}}{f}_{i,j}& j=\mathrm{1,2},...,24,---\left(1\right)\end{array}$

Calculate the average frame of gained as shown in Figure 5, f _i,jwith the correlation coefficient ρ of the average frame of its correspondence position _i,jbe calculated as follows:

$\begin{array}{ccc}{\mathrm{\ρ}}_{i,j}=\mathrm{corr}(f_{i,j},{\stackrel{\‾}{f}}_j)& i=\mathrm{1,2},...,m& j=\mathrm{1,2},...,24,---\left(2\right)\end{array}$

In formula (2), corr () is Calculation of correlation factor function.ρ _i,jvalue shows that more greatly this frame quality is better.

Step 5: according to gait frame image quality judgment criteria, chooses the final cycle expression that top-quality optimal frames forms whole section of gait sequence from each cycle.Particularly, for a jth frame position, its optimal frames for:

$\begin{array}{cc}{f}_j^{*}=\underset{i=1,...,m}{\arg \max }{\mathrm{\ρ}}_{i,j}& j=\mathrm{1,2},...24,---\left(3\right)\end{array}$

If only have one-period in gait sequence, then skip over step 4 and step 5, the final cycle directly using the result of step 3 as this sequence expresses.As shown in Figure 6, express for using the inventive method to comprise from one the gait cycle generated the gait sequence of repeating frame and defective frame.

Above-described embodiment is the present invention's preferably embodiment; but embodiments of the present invention are not restricted to the described embodiments; change, the modification done under other any does not deviate from Spirit Essence of the present invention and principle, substitute, combine, simplify; all should be the substitute mode of equivalence, be included within protection scope of the present invention.

Claims (7)

1. a new gait cycle generation method, is characterized in that, comprise the steps:

Step 1, from pending gait sequence, detect each original gait cycle;

Step 2, set the final cycle express in the frame number k that comprises, this frame number should be less than the minimum value comprising frame number in each original gait cycle of step 1 gained;

Step 3, the frame number set according to step 2, equally spaced extract the representative frame of k frame as this cycle from each original gait cycle;

The quality assessment standard of step 4, definition gait two field picture, and its quality coefficient is calculated to all representative frame obtained by step 3;

Step 5, for the representative frame in same position in all gait cycles, choose the optimal frames of the best frame of picture quality as this position, the optimal frames of each position forms final gait cycle and expresses.

2. gait cycle generation method according to claim 1, it is characterized in that, each original gait cycle is detected according to the cyclical variation of paces width in step 1, the frame comprised between adjacent two paces width maximum value or minimum values is half period, and continuous two semiperiods form a gait cycle.

3. gait cycle generation method according to claim 1, is characterized in that, the minimum value that the numerical value of setting in step 2 should comprise frame number than each the original gait cycle calculated in step 1 is little, and this numerical value must be the integral multiple of 4.

4. gait cycle generation method according to claim 1, it is characterized in that, using 1/4th gait cycles as elementary cell in step 3, each elementary cell is divided at equal intervals with set numerical value in step 2 1/4th, choose from the nearest frame of Along ent as the representative frame of place elementary cell on this position, namely the representative frame of adjacent four elementary cells forms the expression of a complete gait cycle.

5. gait cycle generation method according to claim 1, is characterized in that, the quality coefficient characterizing gait frame image quality quality in step 4 is the related coefficient of the average frame of this frame and its position, and related coefficient is larger, then this frame quality is better; Average frame is by acquisition of averaging to the representative frame being in same position in each cycle.

6. gait cycle generation method according to claim 1, it is characterized in that, according to the frame image quality judgment criteria that step 4 defines in step 5, in comparison gait sequence, all gait cycles are in the quality level of the frame of correspondence position, choose the optimal frames of the best frame of wherein picture quality as this position, the optimal frames on all correspondence positions forms the final gait cycle expression that this gait sequence generates.

7. gait cycle generation method according to claim 1, is characterized in that, the position in step 5 refers to the sequence number of certain representative frame in a current period k representative frame.