CN104408395A - A gesture identifying method and system - Google Patents
A gesture identifying method and system Download PDFInfo
- Publication number
- CN104408395A CN104408395A CN201410290001.4A CN201410290001A CN104408395A CN 104408395 A CN104408395 A CN 104408395A CN 201410290001 A CN201410290001 A CN 201410290001A CN 104408395 A CN104408395 A CN 104408395A
- Authority
- CN
- China
- Prior art keywords
- hand
- user
- face
- gesture
- record
- 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.)
- Pending
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
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/107—Static hand or arm
Abstract
The present invention provides a gesture identifying method and system. Before the gesture of a user is photographed and identified, a position of the user is first determined according to a photographed user image, then according to the position of the user, an angle of a camera apparatus is adjusted to aim at the user; meanwhile, according to a gesture range of the user, a focal length of the camera apparatus is adjusted so as to maximize the user image photographed by the camera apparatus as much as possible. In this way, since the hand of the user is relatively large in the photographed image, the hand is easy to identify and the accuracy rate of identifying is high; and since the camera apparatus would automatically adjust according to the position of the user, the user does not have to specially stand in front of the camera apparatus for operation. Therefore, the gesture identifying method in the present invention is more flexible.
Description
Technical field
The present invention relates to image identification technical field, in particular to a kind of gesture identification method and system.
Background technology
The video equipments such as televisor due to user be all remote viewing, therefore mostly use a teleswitch operation, and along with the intellectuality of product is more and more higher, the mode of operation used a teleswitch can not meet the demand of user.Gesture controls as a kind of novel control mode, is applied to gradually controlling on the products such as televisor.
Comparatively typical Gesture Recognition utilizes camera to shooting image at present, by identifying form and the displacement of the hand of user to the analysis of image, thus determines the gesture of user.Gesture Recognition achieves the conversion to order of the seizure of hand information and hand information.Current gesture identification mainly contains two kinds of recognition method, and a kind of is identify the activities of hand, the different shape mainly pointed, such as, clench one's fists, the sub-gesture of V, OK gesture etc.; Also having a kind of is identify the movement of hand, by identifying that the motion track of hand judges the meaning of gesture, such as, moves left and right palm, moves palm etc. up and down.Activities due to hand needs to identify the details of hand, requires higher comparatively speaking to accuracy of identification, and therefore it implements comparatively difficulty, and discrimination is also undesirable in actual applications.Because this reason, at present in the gesture identification of the products such as televisor, extensive by carrying out gesture identification Application comparison to the motion track of hand.
But also there are some problems by carrying out gesture identification to the motion track of hand in prior art, such as move left and right in the identifying of palm, after user is moved to the left palm, if also need to repeat same action, carry out second time action after palm certainly will be needed to move right, in the process, the move right action of palm of user is also likely identified, that is, there is identification by mistake.And in actual applications, be moved to the left palm and the palm often corresponding different operational order that moves right, if user is moved to the left palm in above process and the palm that moves right is simultaneously identified, this by mistake identification makes user cannot pass on required operational order in this way, thus the equipment of uncontrollable correspondence.
In order to avoid similar maloperation in prior art, often need restriction user being had to each side, but this restriction makes the Consumer's Experience of gesture identification not good.Therefore, need a kind of Gesture Recognition, the mistake identification in gesture identification can be avoided while not limited subscriber.
Summary of the invention
The invention provides a kind of gesture identification method, while not limited subscriber, effectively can avoid the mistake identification in gesture identification.
Concrete, the invention provides a kind of gesture identification method, comprising:
Obtain user images, determine face and the hand of user according to user images;
Judge the face of user and the depth distance of hand, when hand and facial depth distance are greater than the threshold values preset, record is carried out to the position of hand; When hand and facial depth distance are less than the threshold values preset, stop carrying out record to the position of hand;
Determine the motion track of hand according to the position of the hand of record, gesture is identified.
Preferably, gesture identification method of the present invention determines the face of user by carrying out skin color model and eye recognition to user images.
Preferably, gesture identification method of the present invention, by also receiving to the face of user and hand utilizing emitted light signal the light returned from face and the hand of user, obtains the face of user and the depth distance of hand by the flight time of light signal.
Preferably, gesture identification method of the present invention calculates the flight time of light signal according to the phase in-migration of light signal.
Preferably, light signal is launched by gesture identification method of the present invention after high frequency modulated.
Preferably, the threshold values preset described in is determined according to the facial width of user.
Preferably, gesture identification method of the present invention, by calculating the center-of-mass coordinate of hand, carries out record to the center-of-mass coordinate of hand.
Preferably, the center-of-mass coordinate of described calculating hand comprises: choose at least 3 coordinate points from the edge of hand, calculates the polygonal center of gravity formed by these coordinate points, these barycentric coordinates is defined as the center-of-mass coordinate of hand.
Present invention also offers a kind of gesture recognition system, comprise picture pick-up device and gesture identification equipment, picture pick-up device comprises depth survey unit, and gesture identification equipment comprises graphics processing unit, record cell and recognition unit; The user images that described graphics processing unit obtains according to picture pick-up device, determines face and the hand of user; Described depth survey unit is measured the face of user and the depth distance of hand respectively, and judge whether the depth distance of hand and face is greater than the threshold values preset, when hand and facial depth distance are greater than the threshold values preset, record cell carries out record to the position of hand; When hand and facial depth distance are less than the threshold values preset, record cell stops carrying out record to the position of hand; Described recognition unit is according to the motion track of the hand position determination hand of recording unit records, and the motion track according to hand identifies gesture.
Preferably, graphics processing unit determines the face of user by carrying out skin color model and eye recognition to user images.
Preferably, depth survey unit comprises illumination unit, reflected light receiving element, described illumination unit is to the face of user and hand utilizing emitted light signal, reflected light receiving element receives from the face of user and the reflected light of hand, and depth survey unit obtains the face of user and the depth distance of hand by the flight time of light signal.
Preferably, illumination unit is launched after carrying out high frequency modulated to light signal.
Preferably, record cell calculates the center-of-mass coordinate of hand, carries out record to the center-of-mass coordinate of hand.
Preferably, described record cell, by choosing at least 3 coordinate points from the edge of hand, calculates the polygonal center of gravity formed by these coordinate points, these barycentric coordinates is defined as the center-of-mass coordinate of hand.
Preferably, the center-of-mass coordinate of the hand in each two field picture order is connected and obtains the motion track of hand by recognition unit, identifies gesture according to this motion track.
Gesture identification method of the present invention, by the depth distance to user's face and hand, when the depth distance of user's face and hand is greater than threshold values, then the gesture of user is identified, and when the depth distance of user's face and hand is less than threshold values, then do not identify.After adopting this method of gesture identification method of the present invention, when user needs to be controlled relevant device by gesture, only hand need be stretched out forward, make the gesture motion of needs, because the hand distance face of now user is comparatively far away, then this gesture motion can identified, and after user is by hand retraction, because the hand distance face of now user is comparatively near, therefore its action can not be identified.So just effectively can avoid the mistake identification in existing gesture identification method, meanwhile, what measure due to this method is the relative distance of user's face and hand, and user is in office, and where position can operate, therefore can not to the position Constrained of user and restriction.As can be seen here, gesture identification method of the present invention effectively can avoid the mistake identification in gesture identification while not limited subscriber.
Accompanying drawing explanation
Fig. 1 is the schematic diagram of gesture identification method according to an embodiment of the invention;
Fig. 2 is the schematic diagram of gesture recognition system according to an embodiment of the invention.
Embodiment
In order to more clearly understand above-mentioned purpose of the present invention, feature and advantage, below in conjunction with the drawings and specific embodiments, the present invention is further described in detail.
Set forth a lot of detail in the following description so that fully understand the present invention, but the present invention can also adopt other to be different from other modes described here and implement, and therefore, the present invention is not limited to the restriction of following public specific embodiment.
Fig. 1 is the process flow diagram of the gesture identification method according to the embodiment of the present invention.
As shown in Figure 1, gesture identification method can comprise the following steps according to an embodiment of the invention:
Step 102, obtains user images, determines face and the hand of user according to user images;
Concrete, user images can be single picture, also can be continuous print image.To identify the gesture of user due to follow-up, and the gesture of user is a lasting process, therefore for convenience's sake, user images is all generally continuous print image, continuous print image can adopt general continuous shooting to obtain by the first-class equipment of shooting, also can obtain by interval shooting.
The face of user by carrying out skin color model to image, the basis of skin color model can be determined the position of eyes of user, like this, just can determine the face of user accordingly.The face recognition of user out after, remaining flesh tone portion can be defined as hand.In some cases, the colour of skin of the interference such as leg in the colour of skin, can be comprised, in this case can by the position relationship of area of skin color and face, and other details judges, the feature etc. such as pointed.
To the present embodiment be described in detail how be determined by continuous print image face and the hand of user below.
Although user images is continuous print image, the present embodiment, when identifying user's face and hand, needs to identify the user's face in every frame or every a few two field picture and hand.
The face of user and hand carry out identification and can realize by the following method in every two field picture:
Generally, the image that camera obtains is rgb format.R value in rgb format, G value, B value represent red, green and blue value respectively.Generally, it is leading that the colour of skin of human body is all that redness accounts for, how illumination changes is all like this, and in image appearance, the R value being exactly human body complexion is greater than G value and B value, according to this feature, primary screening is carried out to the value of R, G, B, rejects the region that R value is less than G value or B value, only leave qualified color area, if the interference of the color not having other close with the colour of skin, the region of human body complexion so just can be detected substantially.
In order to identify area of skin color more accurately, picture proceeds to detect by the present embodiment on the basis that RGB detects in YCrCb and/or HSV color space.HSV is a color space representing form and aspect, saturation degree and brightness, and the model of this color space corresponds to a conical subset in cylindrical-coordinate system.In hsv color model, each color and its complementary color differ 180 °.Saturation degree S value is from 0 to 1, so the radius of circular cone end face is 1.The color gamut of hsv color model representative is a subset of CIE chromaticity diagram, and in this model, saturation degree be absolutely color, and its purity is generally less than a hundred per cent.YCrCb is the color space representing aberration, and Cr and Cb is red and green difference and difference that is green and blueness, and these two differences of human body have certain limit.The H value utilized in HSV color space, this value is equivalent to the brightness of image, utilizes hue value value can reject some over-exposed errors caused.Such as, the scope of H value can be chosen as 0.1>H>0.01; And the ratio range of Cr/Cb is chosen as 1.1786>Cr/Cb>0.5641, or also Cr and Cb can be calculated respectively, threshold range is chosen as 165>Cr>110,195>Cb>140, but also can adjust according to actual needs.By such detection and screening, the region of human body complexion in image can be detected equally.
The detection of human body complexion above also can be additive method, such as all carry out threshold values judgement by the value of R, G, B of image, the colour of skin can be detected by means of only RGB image, or also first can carry out filtering to the histogram of image, the part meeting features of skin colors is retained after filtering, then carry out second time at YCrCb color space to detect, the human body complexion detected like this is more accurate.
After human body skin tone testing out, continue the eye areas extracting human body according to the shape of area of skin color and positional information.The detection of eye areas has multiple applications, the human-eye positioning methods such as such as conventional Hough transform method, deforming template method, edge feature analytic approach and symmetry transformation method.In embodiments of the present invention, in order to analytic method is convenient, locate to eyes in the following way: the general region that the heart is on the upper side in the picture, position of general human eye, the shape of two other eyes should be similar, namely using circle as the approximate shapes in this region, radius of a circle will differ less, and this is the eye shape of area of skin color.Height is consistent in the horizontal direction for two eye center, again can not be too near on vertical direction, and the region stayed with this is candidate region, namely determines the position of eyes.After the eyes of user are identified, can confirm the face of user, namely the area of skin color at eyes place is exactly the facial zone of user.
In superincumbent Face Detection process, except the colour of skin of face is detected, other exposed positions of user can be detected too, comprise hand, sometimes have leg etc.Because the position of eyes is determined, therefore the position of user face also can be determined, then can utilize the position relationship of each colour of skin block, determine the hand of user.Such as, generally, the hand position of people all can on leg, and the horizontal range of distance eyes is comparatively far away, and the area of hand is also little than the area of leg, according to these conditions, can determine the human body that each colour of skin block is corresponding.In addition, also hand can be identified according to other features of area of skin color, the feature etc. such as pointed.
Except these skin color model methods above, the present embodiment can also use other skin color model methods, such as, simply define the skin color model method etc. of the skin color model method of complexion model, the skin color model method of nonparametric complexion model, the skin color model method of parameter complexion model or the skin color detection algorithm based on adaptive threshold.Simple definition complexion model observes by experiment, the area of skin color in define color space is carried out by definition series of rules, the rule adopted includes the simple linear function of Manual definition, complicated nonlinear function, or is automatically found rule by machine learning.Nonparametric complexion model often plants the skin color probability of color by the colour of skin frequency of occurrences direct estimation in training set, common method has look-up table, Bayes method, SOM method etc., the skin color probability value of each color dot in this class model is independent, and accuracy is high, but lacks generalization ability.Parameter complexion model supposition skin distribution meets the concrete mathematical function of certain class, and determine corresponding parameter by training data, model is assumed that usually obeys single Gaussian distribution, many Gaussian distribution, elliptic systems etc., and this model has generalization ability when training data is insufficient.Skin color detection algorithm based on adaptive threshold is different from the histogram detection method of fixed threshold, can produce corresponding optimum segmentation threshold value for different picture materials.By the observation analysis to skin-color probability distributions histogram (SPDH), 4 clues can be extracted and help find optimal threshold, train an artificial nerve network classifier on this basis to determine optimal threshold.
More than illustrate and be all described based on single-frame images, in actual applications, because every frame picture is too short for interval time, although identify can reach the most accurate effect to the face of user in every two field picture and hand, but operand is larger.In order to reduce operand, the present embodiment also after the face of user and hand identify in certain two field picture, can identify every a few frame, such as, every 1 frame, 2 frames, 5 frames, 10 frames etc. again.The interval of each identification can be identical, also can be different.
Except the face carrying out user to single-frame images except above-mentioned and hard recognition, the present embodiment also carries out skin color model by two frames to continuous print two frame, multiframe or interval, multiframe after carrying out computing again.Due in the process obtaining continuous picture, the position of camera is constant, and thus the background parts of picture is constant, and like this, by carrying out computing to two pictures, such as carrying out subtraction to each pixel effectively can remove background.Picture after such computing is only left the part image of user substantially.Next, continue to use skin color model method above to carry out face recognition to this picture and can reach more accurate recognition effect.Certainly, the consecutive image of user may also be and uses additive method to carry out skin color model, and such as, based on the skin color model method of dynamic complexion model, before and after the method utilizes, the correlativity of frame information identifies the colour of skin.
Step 104, judge the face of user and the depth distance of hand, when depth distance that is facial and hand is greater than the threshold values preset, record is carried out to the motion track of hand, when depth distance that is facial and hand is less than the threshold values preset, stop carrying out record to the motion track of hand;
Face and the hand of step 102 couple user identify, this enforcement of detailed description how to be detected the depth distance of user's face and hand below.
In the present embodiment, the face of user and hand distance detect and can pass through accomplished in many ways.
One method utilizes infrared distance sensor, infrared distance sensor respectively launches a branch of infrared light to user's face and hand, each process forming a reflection after being irradiated to user's face and hand, the signal of infrared distance sensor to reflection receives, then by calculating the data of the mistiming receiving launching and receiving, the distance of user's face and hand and infrared distance sensor can be calculated.
Another kind method is taken user images by two cameras, has certain distance, take from different perspectives respectively to user between two cameras.Due at one time, the face of user and hand position are all identical, and due to the position of two cameras different, therefore the photo photographed has certain difference, according to this difference, the hand of the face of user and the distance of camera and user and the distance of camera can be calculated respectively by methods such as trigonometric functions.
Preferably, the present embodiment, by sending light pulse continuously to the face of user and hand, then with receiving the light returned from face and the hand of user, obtains the face of user and the depth distance of hand by flight (coming and going) time of detecting optical pulses.Concrete, the employing active light detection mode in the present embodiment, utilizes the change of incident optical signal and reflected light signal to carry out range observation.Launch again after light signal is carried out high frequency modulated, such as 100MHz, 200MHz.When the light that the face and hand that receive user return, first filtering is carried out to reflected light, be used for ensureing to only have the light identical with lighting source wavelength to be received.Because the light velocity is determined, like this, by the mistiming of reflected light and back light, the distance of user's face and hand distance luminous point can be calculated respectively.
After the face of user and the depth distance of hand are determined, just can obtain the depth distance between the face of user and hand by simple subtraction.
The determination of threshold values also has multiple method, and can be a concrete numerical value, such as 10 centimetres, 20 centimetres, 30 centimetres etc. also can be relative numerical value, the such as multiple of hand width, the multiple etc. of facial width.Consider the individual difference of different user and the difference at age, use relative numerical value can obtain better effect.Owing to identifying face in a step 102, and the width of human body face ratio is easier to determine, therefore the present embodiment determines threshold values preferably through the width of face, 1 times, 1.5 times, 2 times of such as facial width etc.Certainly, this threshold values also can be user-defined, and user can determine this threshold values according to oneself custom in use.
Above the depth distance between the face of user and hand is determined, again threshold values is determined, can judge whether the depth distance between face and hand is greater than this threshold values so easily.
When the depth distance of user's face and hand is greater than this threshold values, record is carried out to the position of user's hand.Identified by the hand of multiple method to user in step 102.The hand of user is a region in the picture, and these regions form the set of a pixel.Because each pixel in this combination of pixels has the position coordinates determined, be namely therefore that record is carried out to the coordinate of these pixels to the record of user's hand position.
Due to hand position covering is a continuous print region, if recorded the position coordinates of each pixel, needs larger storage space and operation time.In order to address this problem, the present embodiment preferably makes to carry out record to the position of hand in the following method.Method only records an edge coordinate for hand region, and the pixel quantity of such pixel and whole hand region wants much less; Also have a kind of method be the partial coordinates only recorded in the edge coordinate of hand region, the several coordinate record of coordinate or interval at such as flex point place one of them.
When the depth distance of user's face and hand is less than this threshold values, stop carrying out record to the position of user's hand.
Step 106, determines the motion track of hand, mates with the gesture preset according to the position of the hand of record.
Action due to user's hand is continuous print, therefore, regular hour can be continued after the depth distance of user's face and hand is greater than threshold values, during this period, the depth distance of user's face and hand is greater than threshold values all the time, like this, the user's hand position just having some frames goes on record, until after the depth distance of user's face and hand is less than this threshold values.
For the hand position recorded, by hand position can be obtained the motion track of a hand according to time sequencing arrangement.
Preferably, the present embodiment is the motion track that centroid position determines hand by calculating hand.This centroid position can be the center of gravity of the irregular image that hand region is formed, and also can be other point.At step 104, the position of hand records various ways, carries out record comprising to the edge coordinate of hand.The present embodiment can calculate the centroid position of hand by a kind of simple method.In the edge coordinate of hand, choose a uppermost point, a nethermost point, a leftmost point and a rightmost point.When uppermost point is multiple, preferably choose middle point, other three points also can adopt similar mode.Uppermost coordinate points and nethermost some line are formed a straight line, and leftmost point and rightmost some line form another straight line, and this two is in line and has a point of crossing, and namely this point can be used as the centroid position of hand.Like this, the centroid position of the hand of the every two field picture recorded is connected in chronological order, the motion track of hand can be obtained.
The centroid position computing method recorded above are relatively simple, the present embodiment also can use the centroid position of additive method determination hand, at least 3 points at hand edge are such as chosen by interval, these are put line successively and form a polygon, by calculating this polygonal center of gravity, this centre of gravity place is defined as the centroid position of hand.
After the motion track obtaining hand, namely by comparing with the gesture prestored, thus identify the gesture of user.Similar technology has had corresponding prior art, such as Model Matching etc., no longer describes in detail at this.
The gesture identification method of the present embodiment, only has when the depth distance of user's face and hand is greater than threshold values, just identifies the gesture of user, and when the depth distance of user's face and hand is less than threshold values, does not then identify.After adopting in this way, when user needs to be controlled relevant device by gesture, only hand need be stretched out forward, make the gesture motion of needs, because the hand distance face of now user is comparatively far away, then this gesture motion can identified, and after user is by hand retraction, because the hand distance face of now user is comparatively near, therefore its action can not be identified.So just can effectively avoid identifying by mistake.Meanwhile, what measure due to this method is the relative distance of user's face and hand, and user is in office, and where position can operate, therefore can not to the position Constrained of user and restriction.
Figure 2 shows that the schematic diagram of a gesture recognition system of the embodiment of the present invention.
The gesture recognition system 200 of the present embodiment comprises picture pick-up device 210 and gesture identification equipment 220, and wherein, picture pick-up device comprises depth survey unit 212, and described gesture identification equipment 220 comprises graphics processing unit 222, record cell 224 and recognition unit 226.
Picture pick-up device 210 obtains user images by shooting, and then the user images of acquisition is sent to gesture identification equipment 220, graphics processing unit 222 pairs of user images carry out face recognition and hard recognition.The face of user by carrying out skin color model to image, the basis of skin color model can be determined the position of eyes of user, like this, just can determine the face of user accordingly.The face recognition of user out after, remaining flesh tone portion can be defined as hand.In some cases, the colour of skin of the interference such as leg in the colour of skin, can be comprised, in this case can by the position relationship of area of skin color and face, and other details judges, the feature etc. such as pointed.
Preferably, the graphics processing unit 222 of the present embodiment realizes the identification of face to user and hand in the following manner:
Generally, the image that picture pick-up device 210 obtains is rgb format.R value in rgb format, G value, B value represent red, green and blue value respectively.Generally, it is leading that the colour of skin of human body is all that redness accounts for, how illumination changes is all like this, and in image appearance, the R value being exactly human body complexion is greater than G value and B value, according to this feature, primary screening is carried out to the value of R, G, B, rejects the region that R value is less than G value or B value, only leave qualified color area, if the interference of the color not having other close with the colour of skin, the region of human body complexion so just can be detected substantially.
In order to identify area of skin color more accurately, picture proceeds to detect by the present embodiment on the basis that RGB detects in YCrCb and/or HSV color space.HSV is a color space representing form and aspect, saturation degree and brightness, and the model of this color space corresponds to a conical subset in cylindrical-coordinate system.In hsv color model, each color and its complementary color differ 180 °.Saturation degree S value is from 0 to 1, so the radius of circular cone end face is 1.The color gamut of hsv color model representative is a subset of CIE chromaticity diagram, and in this model, saturation degree be absolutely color, and its purity is generally less than a hundred per cent.YCrCb is the color space representing aberration, and Cr and Cb is red and green difference and difference that is green and blueness, and these two differences of human body have certain limit.The H value utilized in HSV color space, this value is equivalent to the brightness of image, utilizes hue value value can reject some over-exposed errors caused.Such as, the scope of H value can be chosen as 0.1>H>0.01; And the ratio range of Cr/Cb is chosen as 1.1786>Cr/Cb>0.5641, or also Cr and Cb can be calculated respectively, threshold range is chosen as 165>Cr>110,195>Cb>140, but also can adjust according to actual needs.By such detection and screening, the region of human body complexion in image can be detected equally.
Graphics processing unit 222 also can use additive method to carry out the colour of skin, such as all carry out threshold values judgement by the value of R, G, B of image, the colour of skin can be detected by means of only RGB image, or also first can carry out filtering to the histogram of image, the part meeting features of skin colors is retained after filtering, then carry out second time at YCrCb color space to detect, the human body complexion detected like this is more accurate.
Graphics processing unit 222 to human body skin tone testing out after, continue the eye areas extracting human body according to the shape of area of skin color and positional information.The detection of eye areas has multiple applications, the human-eye positioning methods such as such as conventional Hough transform method, deforming template method, edge feature analytic approach and symmetry transformation method.In embodiments of the present invention, in order to analytic method is convenient, locate to eyes in the following way: the general region that the heart is on the upper side in the picture, position of general human eye, the shape of two other eyes should be similar, namely using circle as the approximate shapes in this region, radius of a circle will differ less, and this is the eye shape of area of skin color.Height is consistent in the horizontal direction for two eye center, again can not be too near on vertical direction, and the region stayed with this is candidate region, namely determines the position of eyes.After the eyes of user are identified, can confirm the face of user, namely the area of skin color at eyes place is exactly the facial zone of user.
In superincumbent Face Detection process, except the colour of skin of face is detected, other exposed positions of user can be detected too, comprise hand, sometimes have leg etc.Because the position of eyes is determined, therefore the position of user face also can be determined, then can utilize the position relationship of each colour of skin block, determine the hand of user.Such as, generally, the hand position of people all can on leg, and the horizontal range of distance eyes is comparatively far away, and the area of hand is also little than the area of leg, according to these conditions, can determine the human body that each colour of skin block is corresponding.In addition, also hand can be identified according to other features of area of skin color, the feature etc. such as pointed.
More than illustrate and be all described based on single-frame images, in actual applications, because every frame picture is too short for interval time, although identify can reach the most accurate effect to the face of user in every two field picture and hand, but operand is larger.In order to reduce operand, the graphics processing unit 222 of the present embodiment also after the face of user and hand identify in certain two field picture, can identify every a few frame, such as, every 1 frame, 2 frames, 5 frames, 10 frames etc. again.The interval of each identification can be identical, also can be different.
After the face of graphics processing unit 222 couples of users and hand identify, depth survey unit 212 starts to detect the depth distance of user's face and hand.
Depth survey unit 212 can be various structures composition, and the method for the existing face to user of reality and the detection of hand distance is also different in fact.
Depth survey unit 212 can comprise an infrared distance sensor, infrared distance sensor respectively launches a branch of infrared light to user's face and hand, each process forming a reflection after being irradiated to user's face and hand, the signal of infrared distance sensor to reflection receives, like this, depth survey unit 212, by calculating the data of the mistiming receiving launching and receiving, can calculate the distance of user's face and hand and infrared distance sensor.
Depth survey unit 212 also can comprise two cameras being arranged on different azimuth, is taken from different azimuth by two cameras to user images.Due at one time, the face of user and hand position are all identical, and due to the position of two cameras different, therefore the photo photographed has certain difference, according to this difference, depth survey unit 212 can calculate the hand of the face of user and the distance of camera and user and the distance of camera respectively by methods such as trigonometric functions.
Preferably, the depth survey unit 212 of the present embodiment comprises illumination unit, reflected light receiving element and computing unit, illumination unit is launched after light signal is carried out high frequency modulated, and such as adopt the pulsed light that LED or laser diode are launched, this pulse can reach 100MHz.Reflected light receiving element can be a camera lens collecting light, and this camera lens has a bandpass filter, is used for ensureing to only have the light identical with lighting source wavelength just can enter.Because optical imaging system has transparent effect, the scene of different distance is the concentric spherical of each different-diameter, and non-parallel planes, generally need subsequent processing units to correct this error, do not needing, in accurate especially situation, also can not correct.Computing unit is noted down respectively to the phase place that incident light comes and goes between the face of illumination unit and user or hand, then by calculating utilizing emitted light and reflected light relative phase shift relation, and current radiative frequency, calculate the face of user or the distance of hand distance camera unit respectively.After the face of user and the depth distance of hand are determined, depth survey unit 212 just can calculate the depth distance between the face of user and hand.
Record cell 224 is for comparing hand and facial depth distance and a threshold values prestored, when depth distance that is facial and hand is greater than this threshold values, record is carried out to the position of hand, when depth distance that is facial and hand is less than this threshold values, stop carrying out record to the position of hand.
The hand of user is a region in the picture, and these regions form the set of a pixel.Because each pixel in this combination of pixels has the position coordinates determined, be namely therefore that record is carried out to the coordinate of these pixels to the record of user's hand position.Preferably, the record cell 224 of the present embodiment only carries out record to the edge coordinate of hand region, or only records the partial coordinates in the edge coordinate of hand region, the several coordinate record of coordinate or interval at such as flex point place one of them.
Record cell 224 also can calculate the centroid position of hand further, and this centroid position can be the center of gravity of the irregular image that hand region is formed, and also can be other point.In the edge coordinate of hand, choose a uppermost point, a nethermost point, a leftmost point and a rightmost point.When uppermost point is multiple, preferably choose middle point, other three points also can adopt similar mode.Uppermost coordinate points and nethermost some line are formed a straight line, and leftmost point and rightmost some line form another straight line, and this two is in line and has a point of crossing, and namely this point can be used as the centroid position of hand.Record cell 224 also can use the centroid position of additive method determination hand, at least 3 points at hand edge are such as chosen by interval, these being put line successively and form a polygon, by calculating this polygonal center of gravity, this centre of gravity place being defined as the centroid position of hand.
Action due to user's hand is continuous print, therefore, regular hour can be continued after the depth distance of user's face and hand is greater than threshold values, during this period, the depth distance of user's face and hand is greater than threshold values all the time, like this, record cell 224 will continue the user's hand position recording some frames, until the depth distance of user's face and hand is less than this threshold values.
For the hand position recorded, recognition unit 226 is by obtaining the motion track of a hand according to time sequencing arrangement by hand position, then by the motion track of hand and the gesture prestored being compared, thus the gesture of user is identified.。
Preferably, the gesture identification equipment 220 of the present embodiment determines the motion track of hand by the centroid position calculating hand.This centroid position can be the center of gravity of the irregular image that hand region is formed, and also can be other point.In the edge coordinate of hand, choose a uppermost point, a nethermost point, a leftmost point and a rightmost point.When uppermost point is multiple, preferably choose middle point, other three points also can adopt similar mode.Uppermost coordinate points and nethermost some line are formed a straight line, and leftmost point and rightmost some line form another straight line, and this two is in line and has a point of crossing, and namely this point can be used as the centroid position of hand.Like this, the centroid position of the hand of the every two field picture recorded by record cell 224 is connected in chronological order, can obtain the motion track of hand.The gesture identification equipment 220 of the present embodiment also can use the centroid position of additive method determination hand, at least 3 points at hand edge are such as chosen by interval, these are put line successively and form a polygon, by calculating this polygonal center of gravity, this centre of gravity place is defined as the centroid position of hand.After the motion track obtaining hand, gesture identification equipment 220 namely by comparing with the gesture prestored, thus identifies the gesture of user.Similar technology has had corresponding prior art, such as Model Matching etc., no longer describes in detail at this.
The gesture recognition system of the present embodiment, only have when the depth distance of user's face and hand is greater than threshold values, namely only have when hand reaches forward to a certain degree by user, the action of its hand just can by the gesture recognition system identification of the present embodiment, and when the depth distance of user's face and hand is less than threshold values, then do not identify.Adopt the gesture recognition system of the present embodiment, when user needs to be controlled relevant device by gesture, only hand need be stretched out forward, make the gesture motion of needs, because the hand distance face of now user is comparatively far away, then this gesture motion can be identified thus control relevant device, and after user is by hand retraction, because the hand distance face of now user is comparatively near, therefore its action can not be identified.So just can effectively avoid identifying by mistake.
More than be described with reference to the accompanying drawings according to technical scheme of the present invention, gesture identification method of the present invention is according to the hand of user and facial depth distance, judge the action whether current hand motion of user belongs to needs and identify: when hand is reached certain distance forward by user, gesture identification method starts to identify the gesture of user, and after hand is regained by user, gesture identification terminates; User's action after this can not be identified by mistake, until hand stretches out forward by user again.As can be seen here, gesture identification method of the present invention effectively can avoid the mistake identification in gesture identification process, simultaneously, due to gesture identification method of the present invention measure be user's hand and face relative distance, no matter user is in which orientation or the distance of distance picture pick-up device, can be identify thus user be there is no to the restriction of aspect, position by this method.
The foregoing is only the preferred embodiments of the present invention, be not limited to the present invention, for a person skilled in the art, the present invention can have various modifications and variations.Within the spirit and principles in the present invention all, any amendment done, equivalent replacement, improvement etc., all should be included within protection scope of the present invention.
Claims (17)
1. a gesture identification method, is characterized in that, comprising:
Obtain user images, determine face and the hand of user according to user images;
Judge the face of user and the depth distance of hand, when hand and facial depth distance are greater than the threshold values preset, record is carried out to the position of hand; When hand and facial depth distance are less than the threshold values preset, stop carrying out record to the position of hand;
Determine the motion track of hand according to the position of the hand of record, gesture is identified.
2. gesture identification method according to claim 1, is characterized in that: the face determining user by carrying out skin color model and eye recognition to user images.
3. gesture identification method according to claim 1, is characterized in that: the described judgement face of user and the depth distance of hand comprise:
Also receive to the face of user and hand utilizing emitted light signal the light returned from face and the hand of user, obtain the face of user and the depth distance of hand by the flight time of light signal.
4. gesture identification method according to claim 3, is characterized in that: the flight time of described light signal calculates according to the phase in-migration of light signal.
5. gesture identification method according to claim 3, is characterized in that: described light signal is launched after high frequency modulated.
6. gesture identification method according to claim 1, is characterized in that: described in the threshold values that presets determine according to the facial width of user.
7. gesture identification method according to claim 1, is characterized in that, the described position to hand is carried out record and comprised:
Calculate the center-of-mass coordinate of hand, record is carried out to the center-of-mass coordinate of hand.
8. gesture identification method according to claim 7, is characterized in that, the center-of-mass coordinate of described calculating hand comprises:
Choose at least 3 coordinate points from the edge of hand, calculate the polygonal center of gravity formed by these coordinate points, these barycentric coordinates are defined as the center-of-mass coordinate of hand.
9. the gesture identification method according to claim 7 or 8, is characterized in that, the position according to the hand of record determines that the motion track of hand comprises:
The center-of-mass coordinate of the hand in each two field picture order is connected, according to the motion track of this line determination hand.
10. a gesture recognition system, is characterized in that, described gesture recognition system comprises picture pick-up device and gesture identification equipment, it is characterized in that:
Described picture pick-up device comprises depth survey unit, and described gesture identification equipment comprises graphics processing unit, record cell and recognition unit;
The user images that described graphics processing unit obtains according to picture pick-up device, determines face and the hand of user;
Described depth survey unit is measured the face of user and the depth distance of hand respectively, and judge whether the depth distance of hand and face is greater than the threshold values preset, when hand and facial depth distance are greater than the threshold values preset, record cell carries out record to the position of hand; When hand and facial depth distance are less than the threshold values preset, record cell stops carrying out record to the position of hand;
Described recognition unit is according to the motion track of the hand position determination hand of recording unit records, and the motion track according to hand identifies gesture.
11. gesture recognition systems according to claim 10, is characterized in that: described graphics processing unit determines the face of user by carrying out skin color model and eye recognition to user images.
12. gesture recognition systems according to claim 10, is characterized in that:
Depth survey unit comprises illumination unit and reflected light receiving element, described illumination unit is to the face of user and hand utilizing emitted light signal, reflected light receiving element receives from the face of user and the reflected light of hand, obtains the face of user and the depth distance of hand by the flight time of light signal.
13. gesture recognition systems according to claim 12, is characterized in that: described reflected light receiving element determines distance according to reflected light and radiative phase in-migration.
14. gesture recognition systems according to claim 12, is characterized in that: described illumination unit is launched after carrying out high frequency modulated to light signal.
15. gesture recognition systems according to claim 10, is characterized in that: described record cell calculates the center-of-mass coordinate of hand, carries out record to the center-of-mass coordinate of hand.
16. gesture identification methods according to claim 15, it is characterized in that, described record cell, by choosing at least 3 coordinate points from the edge of hand, calculates the polygonal center of gravity formed by these coordinate points, these barycentric coordinates is defined as the center-of-mass coordinate of hand.
17. gesture identification methods according to claim 15 or 16, is characterized in that, the center-of-mass coordinate of the hand in each two field picture order is connected and obtains the motion track of hand by described recognition unit, identifies gesture according to this motion track.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410290001.4A CN104408395A (en) | 2014-06-26 | 2014-06-26 | A gesture identifying method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410290001.4A CN104408395A (en) | 2014-06-26 | 2014-06-26 | A gesture identifying method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN104408395A true CN104408395A (en) | 2015-03-11 |
Family
ID=52646026
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410290001.4A Pending CN104408395A (en) | 2014-06-26 | 2014-06-26 | A gesture identifying method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104408395A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105550655A (en) * | 2015-12-16 | 2016-05-04 | Tcl集团股份有限公司 | Gesture image obtaining device and method |
| WO2017067465A1 (en) * | 2015-10-21 | 2017-04-27 | 华为技术有限公司 | Gesture recognition method and device |
| CN107491717A (en) * | 2016-06-13 | 2017-12-19 | 科大讯飞股份有限公司 | The detection method that cheats at one's exam and device |
| CN107517517A (en) * | 2017-08-25 | 2017-12-26 | 深圳市光域物联科技有限公司 | Multifunctional smart lamp, control method and driving method |
| CN108717326A (en) * | 2018-05-11 | 2018-10-30 | 深圳增强现实技术有限公司 | A kind of anti-tampering gesture identification method and AR glasses based on AR glasses |
| CN109032354A (en) * | 2018-07-24 | 2018-12-18 | Oppo（重庆）智能科技有限公司 | Electronic device and its gesture identification method, computer readable storage medium |
| CN112596605A (en) * | 2020-12-14 | 2021-04-02 | 清华大学 | AR (augmented reality) glasses control method and device, AR glasses and storage medium |
| CN114758208A (en) * | 2022-06-14 | 2022-07-15 | 深圳市海清视讯科技有限公司 | Attendance checking equipment adjusting method and device, electronic equipment and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1218936A (en) * | 1997-09-26 | 1999-06-09 | 松下电器产业株式会社 | Hand gesture identifying device |
| US20090077504A1 (en) * | 2007-09-14 | 2009-03-19 | Matthew Bell | Processing of Gesture-Based User Interactions |
| US20130278493A1 (en) * | 2012-04-24 | 2013-10-24 | Shou-Te Wei | Gesture control method and gesture control device |
| CN103729053A (en) * | 2012-10-16 | 2014-04-16 | 纬创资通股份有限公司 | Gesture control device and method for setting and cancelling gesture operation area in gesture control device |
-
2014
- 2014-06-26 CN CN201410290001.4A patent/CN104408395A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1218936A (en) * | 1997-09-26 | 1999-06-09 | 松下电器产业株式会社 | Hand gesture identifying device |
| US20090077504A1 (en) * | 2007-09-14 | 2009-03-19 | Matthew Bell | Processing of Gesture-Based User Interactions |
| US20130278493A1 (en) * | 2012-04-24 | 2013-10-24 | Shou-Te Wei | Gesture control method and gesture control device |
| CN103729053A (en) * | 2012-10-16 | 2014-04-16 | 纬创资通股份有限公司 | Gesture control device and method for setting and cancelling gesture operation area in gesture control device |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017067465A1 (en) * | 2015-10-21 | 2017-04-27 | 华为技术有限公司 | Gesture recognition method and device |
| US10732724B2 (en) | 2015-10-21 | 2020-08-04 | Huawei Technologies Co., Ltd. | Gesture recognition method and apparatus |
| CN105550655A (en) * | 2015-12-16 | 2016-05-04 | Tcl集团股份有限公司 | Gesture image obtaining device and method |
| CN107491717A (en) * | 2016-06-13 | 2017-12-19 | 科大讯飞股份有限公司 | The detection method that cheats at one's exam and device |
| CN107517517A (en) * | 2017-08-25 | 2017-12-26 | 深圳市光域物联科技有限公司 | Multifunctional smart lamp, control method and driving method |
| CN107517517B (en) * | 2017-08-25 | 2019-08-06 | 深圳市光域物联科技有限公司 | Multifunctional smart lamp, control method and driving method |
| CN108717326A (en) * | 2018-05-11 | 2018-10-30 | 深圳增强现实技术有限公司 | A kind of anti-tampering gesture identification method and AR glasses based on AR glasses |
| CN109032354A (en) * | 2018-07-24 | 2018-12-18 | Oppo（重庆）智能科技有限公司 | Electronic device and its gesture identification method, computer readable storage medium |
| CN109032354B (en) * | 2018-07-24 | 2021-05-07 | Oppo（重庆）智能科技有限公司 | Electronic device, gesture recognition method thereof and computer-readable storage medium |
| CN112596605A (en) * | 2020-12-14 | 2021-04-02 | 清华大学 | AR (augmented reality) glasses control method and device, AR glasses and storage medium |
| CN114758208A (en) * | 2022-06-14 | 2022-07-15 | 深圳市海清视讯科技有限公司 | Attendance checking equipment adjusting method and device, electronic equipment and storage medium |
| CN114758208B (en) * | 2022-06-14 | 2022-09-06 | 深圳市海清视讯科技有限公司 | Attendance checking equipment adjusting method and device, electronic equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104408395A (en) | A gesture identifying method and system | |
| CN104317385A (en) | Gesture identification method and system | |
| CN107169405B (en) | Method and device for living body identification based on binocular camera | |
| JP6858650B2 (en) | Image registration method and system | |
| US8983202B2 (en) | Smile detection systems and methods | |
| JP2021114307A (en) | Information processing device, information processing method, and program | |
| US20200082160A1 (en) | Face recognition module with artificial intelligence models | |
| CN106991654A (en) | Human body beautification method and apparatus and electronic installation based on depth | |
| CN110168562B (en) | Depth-based control method, depth-based control device and electronic device | |
| TW201915831A (en) | System and method for entity recognition | |
| CN106937049A (en) | The processing method of the portrait color based on the depth of field, processing unit and electronic installation | |
| KR100845969B1 (en) | The Extraction method of moving object and the apparatus thereof | |
| JP2000082147A (en) | Method for detecting human face and device therefor and observer tracking display | |
| CN110264493A (en) | A kind of multiple target object tracking method and device under motion state | |
| CN105844242A (en) | Method for detecting skin color in image | |
| CN104392210A (en) | Gesture recognition method | |
| CN108416291A (en) | Face datection recognition methods, device and system | |
| JP6157165B2 (en) | Gaze detection device and imaging device | |
| CN110378946A (en) | Depth map processing method, device and electronic equipment | |
| CN110021029A (en) | A kind of real-time dynamic registration method and storage medium suitable for RGBD-SLAM | |
| CN109359577A (en) | A kind of Complex Background number detection system based on machine learning | |
| KR20160009972A (en) | Iris recognition apparatus for detecting false face image | |
| JP2009086824A (en) | Chin position calculating apparatus, image processing apparatus using it, and chin position calculating method and program | |
| CN112668539A (en) | Biological characteristic acquisition and identification system and method, terminal equipment and storage medium | |
| JP2010237976A (en) | Light source information obtaining device, shading detection device, shading removal device, and those methods and programs |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20150311 |
|
| WD01 | Invention patent application deemed withdrawn after publication |