CN103279188A - Method for operating and controlling PPT in non-contact mode based on Kinect - Google Patents

Abstract

The invention relates to a method for operating and controlling a PPT in a non-contact mode based on Kinect. The method comprises the following steps that the Kinect is accessed to a computer system; the position information and the depth information of a palm are acquired by starting the Kinect to track the palm; the two-dimension plane area where the palm is located at a position where the palm is 1000 mm deep away from the center of the Kinect is divided into several functional arrears, namely a laser pen area, an isolating area and a control area; corresponding PPT operating commands at different positions of each area of the palm are defined; finally corresponding operation is carried out on the PPT by a computer according to gesture control commands. According to the method for operating and controlling the PPT in the non-contact mode based on the Kinect, non-contact operation and control of the demonstration of PPT software is achieved, the degree of demonstration freedom is improved, and operation accuracy is high.

Description

A kind of method based on the contactless property of Kinect operation control PPT

One, technical field

The present invention relates to the method for control PPT demoware in the intelligent meeting, specifically is the method for contactless property operation control PPT demoware.

Two, background technology

In recent years, along with the generally raising of the intelligent degree of meeting, the official meeting of nearly all industry all can use Power Point to assist demonstration.Become the indispensable presentation tool of various official meetings along with PPT, using to the user of this software provides very big convenience, simultaneously a lot of researchists also control PPT and have carried out related development, comprising wireless receiver laser pen, sound control PPT, gesture control PPT.

Though on the one hand above-mentioned achievement in research has been simplified the contact operation between man-machine, and the speechmaker is freed from the operation of contact fully.PPT itself be exactly Microsoft be that purpose is developed to make things convenient for the user, the demonstration of contact operation PPT software has limited operating personnel and has demonstrated degree of freedom in the conference content process.The method sensitivity of existing sound control PPT, gesture control PPT on the other hand, accuracy is relatively poor, can not satisfy speaker's demands of applications.

The Kinect novel 3D body sense picture pick-up device that to be Microsoft release on June 14th, 2010, it can carry out Depth Imaging to scene on every side, and can obtain skeleton information, can be used for the identification of human body limb language.Kinect is with low cost, powerful, itself and software kit do not provide the function of contactless property operation control PPT, use it in the intelligent meeting but its characteristic that has is applicable to fully, realize the purpose of contactless property operation control PPT demoware.

Three, summary of the invention

Problem solved by the invention, the controlling of existing exactly PPT mostly are contact and control, and can not make the speaker disengage operation tool, very flexible; Existing contactless property is controlled the method for PPT such as sound control, some gesture control, and it is low mostly to be sensitivity, and the operation that accuracy is lower can not be satisfied the application needs of speaker in the lecture process.

In order to overcome the above problems, the present invention utilizes shooting and the palm of the hand following function of Kinect camera, and a kind of method based on the contactless property of Kinect operation control PPT is provided, and may further comprise the steps:

A, Kinect is inserted computer system;

B, by starting the tracking of the palm of the hand of Kinect, obtain the positional information of the palm of the hand, and the depth information of the palm of the hand, in the position of the palm of the hand apart from the Kinect depth transducer center 1000mm degree of depth, the two dimensional surface zone at palm of the hand place is carried out the division of functional area, it is divided into laser pen zone, area of isolation, control area, and define the instruction that the corresponding PPT of palm of the hand diverse location in each zone operates, and make following provisions: gauged distance is the distance that is positioned at 1 meter, Kinect dead ahead; Standard flat is to be positioned at the gauged distance place, the plane at the four corner place that the palm of the hand moves; Standard area is palm of the hand image range on Kinect when standard flat moves, and it is defined as the rectangular area; Brandish the center person of being to use when stretching out the right hand in the front, the palm of the hand is brandished position residing in the scope whole, can be under different depth information; Reference area is the center of palm of the hand moving range on the Kinect image, i.e. the zone of waving on the image that produces in the center perpendicular to the Kinect direction of people;

C, computing machine are carried out corresponding operating according to the gesture steering order to PPT.Realized the effect of contactless property operation Automatic Control PPT presentation process, this method has been avoided the research to the gesture track, has simplified the identification difficulty, and improves discrimination.

The invention has the beneficial effects as follows: the speaker can realize the presentation process of contactless property operation control PPT software, the speaker is freed from the operation of contact, improve the degree of freedom of its demonstration, and this method has been avoided the research to the gesture track, simplified the identification difficulty, and improving discrimination, accuracy rate and sensitivity, that the present invention simultaneously has is simple and practical, favorable expandability and the advantage being convenient to popularize.

In standard flat, the normal play of the palm of the hand control PPT that movement can be successful in picture and generation laser pen effect.But in actual applications, speaker's hand is not all the time all in standard flat, so the zone of correspondence on two-dimension picture also can produce difference.The degree of depth moving range of speaker's palm of the hand is fixed, but distance is more far away, the moving range of the palm of the hand on picture can reduce, in order to overcome the above problems, as a further improvement on the present invention, the present invention also is included in the b step, when speaker's palm of the hand is not 1000mm apart from the degree of depth at Kinect depth transducer center, also comprise the palm of the hand is transformed into the palm of the hand apart from the process of the two-dimensional signal of the Kinect depth transducer center 1000mm degree of depth apart from the two-dimensional signal of Kinect depth transducer center different depth position, can select to realize by following steps: at first, adopt least square method to determine the moving range of the palm of the hand under any depth information; Then, adopt the mode of ratio according to depth value the 2-D data of the palm of the hand to be transformed into the palm of the hand apart from the 2-D data of the Kinect depth transducer center 1000mm degree of depth.

The user stands in the center of Kinect in the standard area, control office is sex-limited bigger, in order to make the operator can be in the movement that certain limit is arranged perpendicular to the direction about Kinect, as a further improvement on the present invention, the present invention also comprises and being in the b step, when speaker's the palm of the hand departs from Kinect depth transducer center, comprise that also the palm of the hand two-dimensional signal that will depart from Kinect depth transducer center is corrected to the process of the palm of the hand two-dimensional signal of center, and can realize by following steps: at first, calculate the center of brandishing before and after proofreading and correct apart from difference; Then by difference actual palm of the hand coordinates correction to reference area.Updating formula is $\begin{array}{c}{x}_{\mathrm{now}}=x_{\mathrm{before}}+(x_o-x_{\mathrm{middle}})\\ y_{\mathrm{now}}=y_{\mathrm{before}}+(y_o-y_{\mathrm{middle}})\end{array},$ (x wherein _Middle, y _Middle) for proofreading and correct the preceding center of brandishing, (x ₀, y ₀) be the center of brandishing after proofreading and correct, (x _Before, y _Before) for changing the position of the preceding palm of the hand, (x _Now, y _Now) be the palm of the hand position after the conversion.

In order to reach the effect of determining different operating person identity, as a further improvement on the present invention, before by gesture instruction control PPT operation, also comprise the process for PPT operator identification, this step can realize by the mode of recognition of face, and the recognition of face step comprises that people's face detects and the foundation of face database, determine the process of operator's identity.The foundation of described face database can be adopted principal component analysis (PCA), people's face picture of collecting is converted to the eigenface collection that can represent its key distinction: at first by obtaining the mean value of each pixel, generate average man's face picture of these pictures, then feature people face and average man's face are compared, at last ratio, average man's face and the feature people face collection of all training pictures are stored as database.The process of described definite operator's identity comprises: at first be written into average man's face in the database, eigenface collection and eigenvalue matrix, utilize the camera collection image then, image is handled and the face of conducting oneself detects, the picture of input is projected to the principal component analysis (PCA) subspace, search original training picture at last, find out and have the picture of similar ratio, determine one's identity.The described step of finding out the picture that has similar ratio most adopts Euclidean distance: at first calculate the Euclidean distance of input picture and every training picture, calculated distance value relatively then, distance is more for a short time show more similar, and introducing confidence level formula: $\begin{array}{c}\mathrm{pConfidence}=1.0-\\ \sqrt{\mathrm{DestSql}/\left(\mathrm{float}\right)(\mathrm{nTrainFaces}*\mathrm{nEigens})}/255\end{array},$ Wherein pConfidence is the confidence level of recognition result, and scope is between 0 to 1; NTrainFaces is the number of participating in people's face of training in the database.If pConfidence＞0.5 then think and identify successfully, otherwise recognition result is insincere.

In order to realize determining more accurately the purpose of operator's identity, as a further improvement on the present invention, the process of described definite operator's identity adopts the method for repeatedly sampling analysis, namely within a certain period of time, statistics single recognition credibility surpasses 0.5 result's number of times, and have only when the number of times that is identified as same people and reach output during threshold, this recognition result is just as final recognition result, and the number of times that described output threshold refers to be identified as same people accounts for the ratio of whole identification number of times.

Four, description of drawings

Fig. 1 is operating process synoptic diagram of the present invention.

Fig. 2 is the Kinect outside drawing.

Fig. 3 is palm of the hand three-dimensional position key drawing.

Fig. 4 is the zoning plan of palm of the hand image range on Kinect when moving in the standard flat position.

Fig. 5 is the position that detects the palm of the hand palm of the hand position transition diagram when being positioned at laser pen effect zone.

Fig. 6 is palm of the hand moving range figure during apart from Kinect depth transducer center different depth.

Fig. 7 is the correction principle figure that the palm of the hand two-dimensional signal that will depart from Kinect depth transducer center is corrected to the palm of the hand two-dimensional signal of center.

Fig. 8 is the synoptic diagram of recognition of face step.

Fig. 9 is average man's face figure.

Figure 10 is feature people face collection.

Five, embodiment

Make detailed explanation below in conjunction with accompanying drawing with regard to enforcement of the present invention

Fig. 1 is operating process synoptic diagram of the present invention.As shown in Figure 1, at first Kinect and computing machine are connected to form operating system of the present invention, this system hardware partly is computing machine and Kinect equipment, software section mainly is to realize by the MFC programming under the windows environment, comprise calling and MPC Dram drawing technique of OpenCV, wherein the PPT file can be stored in computing machine in advance, also can download automatically from network by the cloud database.

Fig. 2 is the Kinect outside drawing.As shown in Figure 2: left side camera lens is infrared transmitter, and middle camera lens is general common RGB colour TV camera, and the right camera lens is the 3D depth transducer that infrared C MOS camera constitutes.Sensor generates depth image stream, the reproduction surrounding environment of 3D in real time with the speed of per second 30 frames.The Kinect sensor can obtain image RGB and depth image data simultaneously, supports real-time whole body and bone to follow the tracks of, and can identify a series of action.

By starting the tracking of the palm of the hand of Kinect, application program can obtain the positional information of the palm of the hand in picture in its entirety by the api function of OpenNI, and the depth information of the palm of the hand.As shown in Figure 3: a, b represent the position of palm of the hand position in two dimensional image.C represents the palm of the hand apart from the depth information of depth transducer.This is the three-dimensional data of returning by call back function Hand_Update ().

Be in the two dimensional surface of 1000mm at depth information at first, the present invention is divided into the two large divisions with whole zone: laser pen zone and control area.The laser pen zone mainly is to utilize the position of hand in the space, the position is projected to produce the auxiliary PPT explanation of similar laser effect part on the screen; The control area then is whether to control operation such as page turning or back page turning before the PPT according to the position judgment of the palm of the hand.According to above-mentioned two big dividing region, make following provisions: gauged distance is the distance that is positioned at 1 meter, Kinect dead ahead; Standard flat is to be positioned at the gauged distance place, the plane at the four corner place that the palm of the hand moves; Standard area is palm of the hand image range on Kinect when standard flat moves, and is defined as the rectangular area; Brandish the center person of being to use when stretching out the right hand in the front, the palm of the hand is brandished position residing in the scope whole, can be under different depth information; Reference area is the center of palm of the hand moving range on the Kinect image, i.e. the zone of waving on the image that produces in the center perpendicular to the Kinect direction of people.

Standard area is divided into as shown in Figure 4 zone.As shown in Figure 4: picture in its entirety is divided into laser pen effect zone and the control area two parts all around that are positioned at the center.

As the position that detects the palm of the hand is positioned at laser pen effect zone A, system can be according to palm of the hand position in the drawings, as shown in Figure 5, by formula 1 positional information is transformed into computer position in screen when full frame, and utilizes MFC image programming technique to realize that in screen red luminous point moves the effect that is similar to laser pen. $\mathrm{Px}=\frac{1024}{400}*x;$ $\mathrm{Py}=\frac{768}{300}*y.---\left(1\right).$

In the control area, produce maloperation in order to prevent the palm of the hand at two regional intersections, mark off area of isolation all around in the a-quadrant, the area of isolation effect is: do not produce any operation or laser pen effect if the palm of the hand is positioned at area of isolation.Except at area of isolation, when the position of the palm of the hand " began to show " position in the leftmost side, under intelligent space, intelligent lighting system and the automatic curtain system curtain of can turning off the light automatically cooperated PPT to begin displaying; Also the palm of the hand can be moved to " finishing to show " and close PPT; When the palm of the hand was positioned at the upper left corner or the upper right corner, system can be the PPT page turning automatically; In this external explanation process, the speaker can move to the palm of the hand upside or downside and play or close video at any time; The speaker can also pass through that the palm of the hand is moved to the lower left corner and have a rest in meeting explanation midway, or moves to the lower right corner and stop to have a rest and continue explanation.All can there be certain dormancy time to prevent from repeating causing operation after wherein all functions trigger.The com interface control of last PPT is to realize by the modification to routine among the MSDN.Divide by above-mentioned control area, on standard flat, gesture operation can satisfy PPT speaker's demand substantially.

In standard flat, the normal play of the palm of the hand control PPT that movement can be successful in picture and generation laser pen effect.But in actual applications, speaker's hand is not all the time all in standard flat, so the zone of correspondence on two-dimension picture also can produce difference.The moving range of speaker's palm of the hand is fixed, but distance is more far away, and the moving range of the palm of the hand on picture can reduce, as shown in Figure 6.

In order to address the above problem, according to the depth information of palm of the hand position, utilize the analysis to measurement data that the two-dimensional signal of the palm of the hand in different depth position is transformed in the standard area by the least square method unification.It is identical with control effect in standard area to guarantee that by depth information conversion the speaker makes gesture in the degree of depth arbitrarily.As shown in table 1, measurement data is to brandish hand measured data boundary in the Kinect center at different depth by the tester, and the value in the table is the mean value of measurement data.

The palm of the hand under the table 1 different depth information moves data

The data and the depth information h that can be observed four groups of borders by data in the table are approximated to linear relationship.The present invention obtains the best experimental formula on four groups of borders by least square method according to data: four groups of linear experimental formulas are expressed as f (h)=a+bh.A, the solution of b is such as formula 2,3 shown in: $a=\frac{\mathrm{\Σ}{h}_{i}f{\left(h\right)}_i\mathrm{\Σ}{h}_i-\mathrm{\Σf}{\left(h\right)}_i\mathrm{\Σ}{h}_i^2}{{\left(\mathrm{\Σ}{h}_i\right)}^2-\mathrm{n\Σ}{h_i}^2}---\left(2\right),$ $b=\frac{\mathrm{\Σ}{h}_i\mathrm{\Σf}{\left(h\right)}_i-\mathrm{n\Σ}{h}_{i}f{\left(h\right)}_i}{{\left(\mathrm{\Σ}{h}_i\right)}^2-\mathrm{n\Σ}{h_i}^2}---\left(3\right).$ But must be pointed out and have only when having linear relationship between h and the f (h) that the straight line of match is just meaningful, introduces a parameter for this reason: correlation coefficient r, it is defined as:

(4).When the absolute value of r more near 1 the time, illustrate that linear relationship is more good.Obtain a, b brings formula f (h)=a+bh into and draws best experimental formula.Calculate four groups of formula altogether by table 1, the four groups of coefficients and the related coefficient that calculate are as shown in table 2:

Table 2 least square method result of calculation

By related coefficient as can be known, each group data all has linear dependence.The length and width of palm of the hand moving range subtract the top side and draw and can deduct the leftmost side and lower side by the rightmost side.Can draw the moving range of the palm of the hand and the relation of degree of depth h by last table:

Can calculate by above-mentioned formula, when the degree of depth is 1000mm and the error of actual measured value less than 2px, satisfy the needs of gesture control fully.The moving range of the palm of the hand under any depth information determined that will adopt the mode of ratio according to the degree of depth h 2-D data of the palm of the hand to be transformed in the standard area below, conversion formula 6 is as follows $\begin{array}{c}{x}_{\mathrm{final}}=\frac{640}{-0.100917h+740.8714}*x_{\mathrm{now}}\\ y_{\mathrm{final}}=\frac{480}{-0.093771h+574.5143}*y_{\mathrm{now}}\end{array}---\left(6\right),$ X wherein _Now, y _NowBe the actual position of the palm of the hand, x _Final, y _FinalFor corresponding to the coordinate of the standard area of 640*480 after the conversion.Identical with last joint, by analyzing changing later coordinate, just can finish the palm of the hand to the control of PPT.

The above results is that the center that the speaker stands in Kinect draws, and control office is sex-limited bigger, can be in the movement that certain limit is arranged perpendicular to the direction about Kinect in order to make the user, and present embodiment is done following correction to parameter in the formula 6.Because speaker's the center of brandishing may change, thus at first calculate the center of brandishing before and after proofreading and correct apart from difference, again by difference actual palm of the hand coordinates correction to reference area.The center of brandishing before proofreading and correct can be lifted the settling position that tracks behind the right hand when stipulating that the speaker brings into use and be obtained, and is designated as (x _Middle, y _Middle).And Depth Information Acquistion can be passed through in the center after proofreading and correct, because the center of brandishing after proofreading and correct is exactly the center of palm of the hand moving range in the reference area, is designated as:

$(x_o,y_o)=$

$(\frac{1}{2}(-0.100917h+740.8714),\frac{1}{2}(-0.093771h+574.5143)).$

Go out difference by two center calculation, can proofread and correct, correction principle figure as shown in Figure 7.

Represent by above-mentioned coordinate, can try to achieve (x by following updating formula _Now, y _Now):

$\begin{array}{c}{x}_{\mathrm{now}}=x_{\mathrm{before}}+(x_o-x_{\mathrm{middle}})\\ y_{\mathrm{now}}=y_{\mathrm{before}}+(y_o-y_{\mathrm{middle}})\end{array}---\left(7\right).$ (x wherein _Before, y _Before) for changing the position of the preceding palm of the hand, (x _Now, y _Now) be the palm of the hand position after the conversion, after conversion, be equipped with relative to centre bit and move the purpose that also can realize effect same even recycling formula 6 just can be implemented in the people.

In order to realize the identification for different speaker's identity, before the speaker makes the control PPT that uses gesture, can also comprise the process to speaker's recognition of face.System utilizes the Kinect camera by the method for principal component analysis (PCA) the speechmaker to be carried out recognition of face, determines its name, opens with it corresponding Power point manuscript for it automatically.

Face recognition technology mainly is divided into following three parts: (1) sets up the database of people's face; (2) people's face is detected in real time; (3) compare with face data and the database of current acquisition, obtain recognition result.

It is the prerequisite of face database and people's face contrast identification that people's face detects, and the sorter (haarcascade_frontalface_alt.xml) that the present invention adopts OpenCV to provide carries out people's face and detects.Memory space and raising counting yield in order to increase data all adopt gray level image to calculate.At first the camera by Kinect obtains image, is image transitions gray level image then, and with this gray level image as input and utilize OpenCV sorter and function cvHaarDetectObjects() carry out recognition of face.If obtain human face region then continue these data are handled, detect again otherwise then regain image.The flow process that people's face detects as shown in Figure 8.

After obtaining human face region, set up database according to people's face of gathering.At first facial image is adjusted to fixing dimension, uses histogram equalization then and realize the brightness and contrast that fixes.Obtain at last through pretreated people's face picture.For example, in order to set up database, altogether four-player is carried out the collection of people's face, everyone gathers 20 pictures and the different photos of different people is carried out mark by rename.Adopt the method for principal component analysis (PCA) to set up the face data storehouse on this basis.

The method main thought of principal component analysis (PCA) (being eigenface) is: 80 training pictures collecting are converted to " eigenface " collection (eigenface) that can represent these training picture key distinctions.At first by obtaining the mean value of each pixel, generate " average man's face picture " of these pictures (averageface).Then eigenface and " average man's face " are compared.First eigenface is the difference of topmost face, and second eigenface is second important face's difference, by that analogy ... calculate the key distinction of picture in the training set, and represented every width of cloth training picture with the combination of these " differences ".One the training picture may be following composition: (averageface)+(12.1%of eigenface0) – (25.5%of eigenface1)+(8.0%of eigenface2)+... + (0.0%of eigenface79).The shared ratio of this training picture can be expressed as 12.1 ,-25.5,8.0 ..., 0.0}.Because some eigenface that come the back are picture noises or can too big effect not be arranged to picture, so this ratiometer can only be got preceding 30 by dimensionality reduction to remaining most important part only.At last ratio, average man's face and the eigenface collection of all training pictures are stored as database.Average man's face and eigenface collection such as Fig. 9 and shown in Figure 10.

In the recognition of face part, at first be written into average man's face in the database, eigenface collection and eigenvalue matrix.Utilize the camera collection image then, image is handled and the face of conducting oneself detects.After detecting human face region, according to above facial image being carried out pre-service.Use the cvEigenDecomposite () function of OpenCV, the picture of importing is projected to the principal component analysis (PCA) subspace, to obtain the eigenwert of this photo.Search original training picture at last, find out and have the picture of similar ratio, determine one's identity.

When inquiry has the training picture of the most similar ratio, adopt " Euclidean distance ".At first calculate the Euclidean distance of input picture and every training picture, calculated distance value relatively then, apart from more for a short time show more similar.Apart from computing formula as shown in Equation 8.

$\mathrm{DestSql}=\underset{j=1....80}{\min }\underset{i=0}{\overset{N=\mathrm{nEigens}-1}{\mathrm{\Σ}}}\left(\mathrm{TestFace}\right[i]-{\mathrm{TrainFace}}_j{\left[i\right]}^2)---\left(8\right).$ Wherein TestFace is the ratio of detected image, TrainFace _jIt is the ratio of j training image.NEigens is the number of eigenface.By above-mentioned result of calculation, can find the minimum picture of distance, but can not assert this picture with detected be a people.In order to represent the confidence level of recognition result, introduce confidence level computing formula 9:

$\begin{array}{c}\mathrm{pConfidence}=1.0-\\ \sqrt{\mathrm{DestSql}/\left(\mathrm{float}\right)(\mathrm{nTrainFaces}*\mathrm{nEigens})}/255\end{array}---\left(9\right).$ Wherein pConfidence is the confidence level of recognition result, and scope is between 0 to 1; NTrainFaces is the number of participating in people's face of training in the database.If pConfidence＞0.5 then think and identify successfully, otherwise recognition result is insincere.

In order further to improve the accuracy of recognition result, eliminate the interference that the result is caused because of some accidentalia in the identifying, system adopts the method for repeatedly sampling analysis, namely within a certain period of time, statistics single recognition credibility surpasses 0.5 result's number of times, and have only when the number of times that is identified as same people reaches output threshold (number of times that namely is identified as same people accounts for the ratio of whole identification number of times), this recognition result is just as final recognition result.

According to above recognition result, system can find corresponding PPT file and open preparation for it and play, and wherein the PPT file can be also can downloading automatically from network by the cloud database of storing in advance.Behind the File Open, the hand that system starts the people of Kinect carries out motion capture, determines that according to dynamic data corresponding actions controls PPT at last.

Claims (10)

1. method based on the contactless property of Kinect operation control PPT is characterized in that may further comprise the steps:

A, Kinect is inserted computer system;

B, by starting the tracking of the palm of the hand of Kinect, obtain the positional information of the palm of the hand, and the depth information of the palm of the hand, in the position of the palm of the hand apart from the Kinect depth transducer center 1000mm degree of depth, the two dimensional surface zone at palm of the hand place is carried out the division of functional area, it is divided into laser pen zone, area of isolation, control area, and define the instruction that the corresponding PPT of palm of the hand diverse location in each zone operates, and make following provisions: gauged distance is the distance that is positioned at 1 meter, Kinect dead ahead; Standard flat is to be positioned at the gauged distance place, the plane at the four corner place that the palm of the hand moves; Standard area is palm of the hand image range on Kinect when standard flat moves, and it is defined as the rectangular area; Brandish the center person of being to use when stretching out the right hand in the front, the palm of the hand is brandished position residing in the scope whole, can be under different depth information; Reference area is the center of palm of the hand moving range on the Kinect image, i.e. the zone of waving on the image that produces in the center perpendicular to the Kinect direction of people;

C, computing machine are carried out corresponding operating according to the gesture steering order to PPT.

2. the method based on the contactless property of Kinect operation control PPT according to claim 1, it is characterized in that in the b step, when speaker's palm of the hand is not 1000mm apart from the degree of depth at Kinect depth transducer center, also comprise the palm of the hand is transformed into the palm of the hand apart from the process of the two-dimensional signal of the Kinect depth transducer center 1000mm degree of depth apart from the two-dimensional signal of Kinect depth transducer center different depth position.

3. the method based on the contactless property of Kinect operation control PPT according to claim 2, it is characterized in that describedly the palm of the hand is transformed into the palm of the hand in the two-dimensional signal of different depth position can realizes by following steps apart from the process of the two-dimensional signal of the Kinect center 1000mm degree of depth: at first, adopt least square method to determine the moving range of the palm of the hand under depth information arbitrarily; Then, adopt the mode of ratio according to depth value the 2-D data of the palm of the hand to be transformed into the palm of the hand apart from the 2-D data of the Kinect center 1000mm degree of depth.

4. according to any described method based on the contactless property of Kinect operation control PPT of claim 1-3, it is characterized in that in the b step, when speaker's the palm of the hand departs from Kinect depth transducer center, comprise that also the palm of the hand two-dimensional signal that will depart from Kinect depth transducer center is corrected to the process of the palm of the hand two-dimensional signal of center.

5. the method based on the contactless property of Kinect operation control PPT according to claim 4, it is characterized in that the process that the described palm of the hand two-dimensional signal that will depart from Kinect depth transducer center is corrected to the palm of the hand two-dimensional signal of center can realize by following steps: at first, calculate the center of brandishing before and after proofreading and correct apart from difference; Then by difference actual palm of the hand coordinates correction to reference area, updating formula is $\begin{array}{c}{x}_{\mathrm{now}}=x_{\mathrm{before}}+(x_o-x_{\mathrm{middle}})\\ y_{\mathrm{now}}=y_{\mathrm{befor}}+(y_o-y_{\mathrm{middle}}),\end{array}$ (x wherein _Middle, y _Middle) for proofreading and correct the preceding center of brandishing, (x ₀, y ₀) be the center of brandishing after proofreading and correct, (x _Before, y _Before) for changing the position of the preceding palm of the hand, (x _Now, y _Now) be the palm of the hand position after the conversion.

6. according to any described method based on the contactless property of Kinect operation control PPT in the claim 1,2,4,5, it is characterized in that by before the gesture instruction control PPT operation, also comprise the process for PPT operator identification, and realize that by recognition of face the recognition of face step comprises that people's face detects and the foundation of face database, determine the process of speaker's identity.

7. the method based on the contactless property of Kinect operation control PPT according to claim 6, principal component analysis (PCA) is adopted in the foundation that it is characterized in that described face database, people's face picture of collecting is converted to the eigenface collection that can represent its key distinction: at first by obtaining the mean value of each pixel, generate average man's face picture of these pictures, then feature people face and average man's face are compared, at last ratio, average man's face and the feature people face collection of all training pictures are stored as database.

8. the method based on the contactless property of Kinect operation control PPT according to claim 6, the process that it is characterized in that described definite operator's identity comprises: at first be written into average man's face in the database, eigenface collection and eigenvalue matrix, utilize the camera collection image then, image is handled and the face of conducting oneself detects, the picture of input is projected to the principal component analysis (PCA) subspace, search original training picture at last, find out and have the picture of similar ratio, determine one's identity.

9. the method based on the contactless property of Kinect operation control PPT according to claim 8, it is characterized in that the described step of finding out the picture that has similar ratio most adopts Euclidean distance: the Euclidean distance of at first calculating input picture and every training picture, compare the calculated distance value then, distance is more for a short time to be shown more similarly, and introduces the confidence level formula:

$\mathrm{pConfidence}=1.0-$

$\sqrt{\mathrm{DestSql}/\left(\mathrm{float}\right)(\mathrm{nTrainFaces}*\mathrm{nEigens})}/255,$ Wherein pConfidence is the confidence level of recognition result, and scope is between 0 to 1; NTrainFaces is the number of participating in people's face of training in the database.If then thinks and identify successfully, otherwise recognition result is insincere in pConfidence＞0.5.

10. the method based on the contactless property of Kinect operation control PPT according to claim 9, the process that it is characterized in that described definite operator's identity adopts the method for repeatedly sampling analysis, namely within a certain period of time, statistics single recognition credibility surpasses 0.5 result's number of times, and have only when the number of times that is identified as same people reaches the output threshold, this recognition result is just as final recognition result, and the number of times that described output threshold namely is identified as same people accounts for the ratio of whole identification number of times.

Images