CN102467658A - Identification method for smiling face and device thereof and mobile terminal thereof - Google Patents
Identification method for smiling face and device thereof and mobile terminal thereof Download PDFInfo
- Publication number
- CN102467658A CN102467658A CN2010105519280A CN201010551928A CN102467658A CN 102467658 A CN102467658 A CN 102467658A CN 2010105519280 A CN2010105519280 A CN 2010105519280A CN 201010551928 A CN201010551928 A CN 201010551928A CN 102467658 A CN102467658 A CN 102467658A
- Authority
- CN
- China
- Prior art keywords
- smiling face
- target image
- image
- characteristic
- eigenwert
- 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
Images
Landscapes
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses an identification method for a smiling face and a device thereof and a mobile terminal thereof, wherein the method comprises the following steps: selecting characteristic values by the Adaboost algorithm according to the set smiling face recognition rate; preprocessing the target image, wherein the target image is to be judged whether it is a smiling image, and the preprocessing comprises geometrical normalization processing and illumination normalization processing; calculating the convolution template and the partial characteristic points of the processed target image according to the characteristic values; obtaining characteristic vectors by performing the convolution calculation according to the partial characteristic points and the convolution template; sending the characteristic vectors to a classifier to be matched to judge whether the target image is a smiling face image. The identification method for the smiling face and the device thereof and the mobile terminal thereof provided by the invention can be implemented based on the fast partial Gabor filtering and the Adaboost smiling face identification technology to overcome the shortcomings existing in the prior smiling face identification technology, and improve the feature extraction speed, and reduce storage amount so as to be suitable for being applied on an embedded platform with limited resources.
Description
Technical field
The present invention relates to image processing field, relate in particular to a kind of smiling face's recognition methods, device and portable terminal.
Background technology
In recent years; Along with the development of computer vision technique and the promotion of mobile multimedia application need for a business; Smiling face's recognition technology towards Embedded Application has obtained very big development; As the mobile phone smiling face take pictures, application scenarios such as digital camera smiling face shutter, mobile phone smiling face seizure, human face expression analysis, all need smiling face's target be discerned, the experience level of these application and the performance of smiling face's recognition methods are closely related.
But the inventor finds that in the process of embodiment of the present invention all there is obvious defects in existing smiling face's recognition technology;
Present smiling face's recognition technology mainly contains two types; One type is to comprise: the Eigenfaces technology that Turk and Pentland propose, the Fisherface technology that Belhumeur etc. propose, the Gabor feature technology that Lyons and Zhang use etc.; Though the recognition performance of these smiling face's recognition technologies is pretty good; But calculate too complicacy, resource consumption is high, is unfavorable for using at the embedded platform of resource-constraineds such as portable terminal; The another kind of smiling face's recognition technology that is based on PHOG, this technological algorithm fast operation, resource occupation is few, but accuracy of identification is very limited.
Summary of the invention
Embodiment of the invention technical matters to be solved is; A kind of smiling face's recognition methods, device and portable terminal are provided; To the deficiency that exists in existing smiling face's recognition technology, a kind of new smiling face's recognition technology based on quick local Gabor filtering and Adaboost is proposed, improve the speed of feature extraction; Reduce memory space, make on its embedded platform that is adapted at resource-constrained and use.
In order to solve the problems of the technologies described above, the embodiment of the invention provides a kind of smiling face's recognition methods, comprising:
Smiling face's discrimination according to setting utilizes the Adaboost algorithm to select eigenwert;
Target image is carried out pre-service; Said target image judges for waiting whether it is smiling face's image; Said pre-service comprises that geometrical normalization is handled and unitary of illumination is handled;
According to said eigenwert, calculate the convolution template and the local unique point of said pretreated target image;
Carry out convolution algorithm according to said local feature point and convolution template, obtain proper vector;
Said proper vector is delivered to sorter mate, judge whether said target image is smiling face's image.
Wherein, said smiling face's discrimination according to setting utilizes the Adaboost algorithm to select before the eigenwert, also comprises training classifier; Said training classifier comprises:
Make up the image pattern storehouse, comprise smiling face's image and non-smiling face's image in the said image pattern storehouse;
According to the Gabor algorithm all images in the image pattern storehouse is carried out filtering, extract eigenwert;
Utilize the Adaboost algorithm to select characteristic, make up two-stage classification device at least.
Wherein, the convolution template of said target image is:
X '=xcos θ+ysin θ wherein, y '=-xsin θ+ycos θ.
Wherein, the convolution template of the said target image of said calculating comprises:
With bank of filters said target image is carried out the Gabor feature ordering, the big figure of generating feature;
Sampling rate y according to setting samples to the big figure of said characteristic, and arranges formation one-dimensional characteristic vector β in order;
Calculate the yardstick λ and the direction θ of said convolution template, wherein
K=1,2,3.......N, W are that the big figure of characteristic is wide, and w is that target image is wide, and h is that target image is high, and y is a sampling rate, β
kBe the eigenwert of utilizing the Adaboost algorithm to select.
Wherein, the local feature point that calculates said target image comprises:
Unique point coordinate in the calculating target image (x, y), wherein,
x
k=((β
k%W
1The %w of) * y),
Wherein W1 is wide for the big figure of characteristic after sampling, and W is that the big figure of characteristic is wide, and w is that target image is wide, and h is that target image is high, and y is a sampling rate, β
kBe the eigenwert of utilizing the Adaboost algorithm to select.
Accordingly, the present invention also provides a kind of smiling face's recognition device, comprising:
Eigenwert is selected module, is used for utilizing the Adaboost algorithm to select eigenwert according to smiling face's discrimination of setting;
Pre-processing module is used for target image is carried out pre-service, and said target image judges for waiting whether it is smiling face's image; Said pre-service comprises that geometrical normalization is handled and unitary of illumination is handled;
The calculation process module; Be used for selecting the selected eigenwert of module according to said eigenwert; Calculate the convolution template and the local unique point of the target image after said pre-processing module is handled, and carry out convolution algorithm, obtain proper vector according to said local feature point and convolution template;
Characteristic matching module is used for that the proper vector that said calculation process module obtains is delivered to sorter and matees, and judges whether said target image is smiling face's image.
Wherein, said device also comprises: the training sort module, be used to make up the image pattern storehouse, and according to the Gabor algorithm all images in the image pattern storehouse is carried out filtering, extract eigenwert, and utilize the Adaboost algorithm to select characteristic, make up two-stage classification device at least;
Comprise smiling face's image and non-smiling face's image in the said image pattern storehouse.
Wherein, the convolution template of said target image is:
X '=xcos θ+ysin θ wherein, y '=-xsin θ+ycos θ.
Wherein, said calculation process module comprises:
The filtering sampling unit is used for bank of filters said target image being carried out the Gabor feature ordering, the big figure of generating feature; And the big figure of said characteristic is sampled, and arrange in order and constitute one-dimensional characteristic vector β according to the sampling rate y that sets;
The template computing unit is used to calculate the yardstick λ and the direction θ of said convolution template, wherein
K=1,2,3.......N, W are that the big figure of characteristic is wide, and w is that target image is wide, and h is that target image is high, and y is a sampling rate, β
kBe the eigenwert of utilizing the Adaboost algorithm to select;
The coordinate Calculation unit, be used for calculating target image unique point coordinate (x, y); X wherein
k=((β
k%W
1The %w of) * y),
Wherein W1 is wide for the big figure of characteristic after sampling, and W is that the big figure of characteristic is wide, and w is that target image is wide, and h is that target image is high, and y is a sampling rate, β
kBe the eigenwert of utilizing the Adaboost algorithm to select;
The convolution algorithm unit, the local feature point that convolution template that is used for said template computing unit is calculated and said coordinate Calculation unit calculate carries out convolution algorithm, obtains proper vector.
Accordingly, the present invention also provides a kind of portable terminal, has adopted aforementioned smiling face's recognition device in this portable terminal.
Embodiment of the present invention embodiment; To the deficiency that exists in existing smiling face's recognition technology; A kind of new smiling face's recognition technology based on quick local Gabor filtering and Adaboost is proposed; Improved the speed of feature extraction, reduced memory space, made on its embedded platform that is adapted at resource-constrained and use.
Description of drawings
In order to be illustrated more clearly in the embodiment of the invention or technical scheme of the prior art; To do to introduce simply to the accompanying drawing of required use in embodiment or the description of the Prior Art below; Obviously, the accompanying drawing in describing below only is some embodiments of the present invention, for those of ordinary skills; Under the prerequisite of not paying creative work property, can also obtain other accompanying drawing according to these accompanying drawings.
Fig. 1 is smiling face's recognition methods first embodiment schematic flow sheet provided by the invention;
Fig. 2 is smiling face's recognition methods second embodiment schematic flow sheet provided by the invention;
Fig. 3 is local feature point and the schematic flow sheet of convolution template that calculates in smiling face's recognition methods provided by the invention in the target image;
Fig. 4 is an ordering synoptic diagram of selecting 3 * 8 bank of filters for use provided by the invention;
Fig. 5 is smiling face's recognition device first embodiment schematic flow sheet provided by the invention;
Fig. 6 is smiling face's recognition device second embodiment schematic flow sheet provided by the invention.
Embodiment
To combine the accompanying drawing in the embodiment of the invention below, the technical scheme in the embodiment of the invention is carried out clear, intactly description, obviously, described embodiment only is the present invention's part embodiment, rather than whole embodiment.Based on the embodiment among the present invention, those of ordinary skills are not making the every other embodiment that is obtained under the creative work prerequisite, all belong to the scope of the present invention's protection.
Referring to Fig. 1, be smiling face's recognition methods first embodiment schematic flow sheet provided by the invention, during each is implemented in the present invention, possibly relate to Gabor function and Adaboost algorithm.Wherein the Gabor function belongs to the windowing Fourier transform, can on frequency domain different scale, different directions, extract relevant characteristic, because the biological agent of Gabor function and human eye is similar, so often as texture recognition.The Adaboost algorithm is a kind of iterative algorithm, and its core concept is to the different sorter (ground floor sorter) of same training set training, gathers these Weak Classifiers then, constitutes a stronger sorter (second layer sorter).Smiling face's recognition methods that present embodiment provides is as shown in Figure 1, comprising:
At step S100, the smiling face's discrimination according to setting utilizes the Adaboost algorithm to select eigenwert.
At step S101, target image is carried out pre-service; Said target image judges for waiting whether it is smiling face's image; Said pre-service comprises that geometrical normalization is handled and unitary of illumination is handled.
At step S102,, calculate the convolution template and the local unique point of said pretreated target image according to said eigenwert.
At step S103, carry out convolution algorithm according to said local feature point and convolution template, obtain proper vector.
At step S104, said proper vector is delivered to sorter mate, judge whether said target image is smiling face's image.
Smiling face's recognition methods that embodiment of the present invention embodiment provides; To the deficiency that exists in existing smiling face's recognition technology; A kind of new smiling face's recognition technology based on quick local filtering and Adaboost is proposed; Improved the speed of feature extraction, reduced memory space, made on its embedded platform that is adapted at resource-constrained and use.
Referring to Fig. 2, be smiling face's recognition methods second embodiment schematic flow sheet provided by the invention, in the present embodiment, with the flow process of this smiling face's recognition methods of more detailed description.As shown in Figure 2, this smiling face's recognition methods comprises:
At step S200, make up the image pattern storehouse, comprise smiling face's image and non-smiling face's image in the said image pattern storehouse.
At step S201, according to the Gabor algorithm all images in the image pattern storehouse is carried out filtering, extract eigenwert.More concrete, in smiling face's recognition methods that for example present embodiment provides, the bank of filters of employing 3 * 8 can reach the highest precision, then adopts this bank of filters of 3 * 8 according to the Gabor algorithm all images in the image pattern storehouse to be carried out filtering, extracts eigenwert.
At step S202, utilize the Adaboost algorithm to select characteristic, make up two-stage classification device at least, to satisfy different application environments.More concrete, the selection of sorter: ground floor can be used a not too strict sorter (for example adopting 200 positive samples, 400 results that the negative sample training draws); The second layer is with a strict a little sorter (for example adopting 500 positive samples, 1000 results that the negative sample training draws).
Above-mentioned steps is for combining the process of overall Gabor and Adaboost structural classification device in the present invention's enforcement, for presetting step.
At step S203, the smiling face's discrimination according to setting utilizes the Adaboost algorithm to select eigenwert.More concrete, the user can set smiling face's discrimination that oneself requires, and certainly, it is high more that the user sets discrimination, and the eigenwert that Adaboost selects is also just many more, and the resource that the follow-up smiling face of being takies when discerning is just many more, and the speed of identification is just slow more.So the user need select balance on smiling face's discrimination and recognition time.
Preferably, smiling face's discrimination is adjustable in the present embodiment, because misclassification rate is generally lower under the simple situation of background, can turn down smiling face's discrimination this moment, obtained recognition speed faster; Misclassification rate is generally higher under the background complicated situation, and can heighten smiling face's discrimination this moment, has obtained smiling face's identification more accurately.
At step S204, target image is carried out pre-service; Said target image judges for waiting whether it is smiling face's image.Said pre-service comprises that geometrical normalization is handled and unitary of illumination is handled.More concrete, Gabor depends on gradation of image distribution spatially in itself, after obtaining target image, also need carry out a series of processing to it, has reached position correction, the normalized purpose of light.The pure expression zone of pretreated desirable output makes it have unified size, shape and normalized brightness value, and removes the influence of light and intensity of illumination.
At step S205,, calculate the convolution template and the local unique point of pretreated target image according to said eigenwert.More concrete, it is a most important link during the smiling face discerns that smiling face's proper vector is extracted, and present smiling face's recognition technology is not suitable on the embedded platform of resource-constrained, using; Be exactly because the operand that smiling face's proper vector is extracted is excessive, for solving problem, the present invention adopts and extracts smiling face's Gabor characteristic based on the feature extracting method of Gabor wavelet transformation; Though the Gabor characteristic has recognition performance preferably; But its classical Gabor algorithm computation amount and memory requirements are very big, and this is unfavorable for the realization of embedded systems such as mobile phone, in order to improve this present situation; The present invention has designed quick local Gabor filtering algorithm (FL_Gabor; Fast and Local_Gabor), utilizes the FL_Gabor algorithm, the eigenwert that chooses according to Adaboost; The local feature point in the calculating target image and the convolution template of target image, its detailed process is as shown in Figure 3:
At step S300, with bank of filters said target image is carried out the Gabor feature ordering, the big figure of generating feature.More concrete, to target image Gabor feature ordering, for example the present invention selects 3 * 8 bank of filters for use, and the ordering synoptic diagram is as shown in Figure 4, each grid size be 48 * 64 (size that is target image is that w * h), then the length and width of the big figure of characteristic are calculated as follows:
W＝w×8＝48×8＝384；H＝h×3＝64×3＝192；
At step S301, according to the sampling rate y that sets the big figure of said characteristic is sampled, and arrange formation one-dimensional characteristic vector β in order.More concrete, suppose that sampling rate y of the present invention is 3, big or small W1 * H1 of the big figure of sampling back characteristic calculates as follows:
W1＝W/y＝384/3＝128；H1＝H/y＝192/3＝64
At step S302; Calculate the convolution template of said target image; The general use volume product module plate of target image is
x '=xcos θ+ysin θ wherein, y '=-xsin θ+ycos θ.In this step, as long as calculate the yardstick λ and the direction θ of target image convolution template, definite this template that just can be unique.In the present invention,
K=1,2,3.......N, W are that the big figure of characteristic is wide, and w is that target image is wide, and h is that target image is high, and y is a sampling rate, β
kBe the eigenwert of utilizing the Adaboost algorithm to select.
At step S303, calculate said target image the local feature point (x, y) more concrete, in the present invention, unique point coordinate in the target image (x, y) in, x
k=((β
k%W
1The %w of) * y),
Wherein W1 is wide for the big figure of characteristic after sampling, and W is that the big figure of characteristic is wide, and w is that target image is wide, and h is that target image is high, and y is a sampling rate, β
kBe the eigenwert of utilizing the Adaboost algorithm to select.
Need to prove that the execution of step S302 and step S303 in no particular order.So far, we utilize the FL_Gabor algorithm, according to the eigenwert that Adaboost chooses, calculate the convolution template except local feature point in the target image and target image.Get back to Fig. 2:
At step S206, utilize the FL_Gabor algorithm that said local feature point and convolution template are carried out convolution algorithm, obtain proper vector.Because this step is only carried out convolution algorithm to local unique point and convolution template, with respect to existing smiling face's recognition technology, its operand reduces greatly.For example,, have 60000 points,, then need 60000 points and overall template be carried out convolution algorithm according to prior art to one 200 * 300 image.And in the present invention; The same image that is directed against 200 * 300 is selected 3 * 8 bank of filters equally for use, and the local feature point that obtains according to FL Gabor algorithm computation possibly have only 20; Then only need the convolution template of these 20 points and target image be carried out convolution algorithm gets final product; With respect to before the convolution algorithm of 60000 points, difficulty and operand be reduction greatly all, required time of computing is much less very naturally.
At step S207, proper vector is delivered to sorter mate, judge whether said target image is smiling face's image, and the result of output judgement.
Smiling face's recognition methods that embodiment of the present invention embodiment provides; To the deficiency that exists in existing smiling face's recognition technology; A kind of new smiling face's recognition technology based on FL_Gabor algorithm and Adaboost is proposed; Improved the speed of feature extraction, reduced memory space, made on its embedded platform that is adapted at resource-constrained and use.
Referring to Fig. 5, be smiling face's recognition device first example structure synoptic diagram provided by the invention, as shown in Figure 5, this device comprises:
Eigenwert is selected module 1, is used for utilizing the Adaboost algorithm to select eigenwert according to smiling face's discrimination of setting.
Smiling face's recognition device that embodiment of the present invention embodiment provides; To the deficiency that exists in existing smiling face's recognition technology; A kind of new smiling face's recognition technology based on quick local filtering and Adaboost is proposed; Improved the speed of feature extraction, reduced memory space, made on its embedded platform that is adapted at resource-constrained and use.Smiling face's recognition device that present embodiment provides is particularly suitable on mobile phone, using.
Referring to Fig. 6, be smiling face's recognition device second example structure synoptic diagram provided by the invention, in the present embodiment, with the 26S Proteasome Structure and Function of this each module of smiling face's recognition device of more detailed description.As shown in Figure 6, this device comprises:
More concrete; For example smiling face's recognition device of providing of present embodiment can adopt 3 * 8 bank of filters at most; Then train sort module 5 to adopt this bank of filters of 3 * 8 all images in the image pattern storehouse to be carried out filtering, extract eigenwert according to the Gabor algorithm.Then, training sort module 5 utilizes the Adaboost algorithm to select characteristic, makes up two-stage classification device at least, to satisfy different application environments.Selection for sorter: ground floor can be used a not too strict sorter (for example adopting 200 positive samples, 400 results that the negative sample training draws); The second layer is with a strict a little sorter (for example adopting 500 positive samples, 1000 results that the negative sample training draws).
Eigenwert is selected module 1, is used for utilizing the Adaboost algorithm to select eigenwert according to smiling face's discrimination of setting.More concrete, the user can set smiling face's discrimination that oneself requires, certainly; It is high more that the user sets discrimination; The eigenwert that Adaboost selects is also just many more, and the resource that this smiling face's recognition device takies in the follow-up smiling face's of being identification is just many more, and the speed of identification is just slow more.So the user need select balance on smiling face's discrimination and recognition time.
Preferably, smiling face's discrimination is adjustable in the present embodiment, because misclassification rate is generally lower under the simple situation of background, can turn down smiling face's discrimination this moment, obtained recognition speed faster; Misclassification rate is generally higher under the background complicated situation, and can heighten smiling face's discrimination this moment, has obtained smiling face's identification more accurately.
It is a most important link during the smiling face discerns that the face proper vector is extracted; Present smiling face's recognition technology is not suitable on the embedded platform of resource-constrained, using, and is exactly because the operand that smiling face's proper vector is extracted is excessive, for solving problem; Calculation process module 3 of the present invention adopts extracts smiling face's Gabor characteristic based on the feature extracting method of Gabor wavelet transformation; Though the Gabor characteristic has recognition performance preferably, its classical Gabor algorithm computation amount and memory requirements are very big, and this is unfavorable for the realization of embedded systems such as mobile phone; In order to improve this present situation; Calculation process module 3 of the present invention is utilized the FL_Gabor algorithm, according to the eigenwert that Adaboost chooses, and the local feature point in the calculating target image and the convolution template of target image.
More concrete, this calculation process module 3 comprises:
More concrete; For example the present invention selects 3 * 8 bank of filters for use; The ordering synoptic diagram is as shown in Figure 4; Each grid size is 48 * 64, and (size that is target image is w * h), and then the 31 pairs of target images in filtering sampling unit length and width of carrying out the big figure of characteristic behind the Gabor feature ordering are calculated as: W=w * 8=48 * 8=384; H=h * 3=64 * 3=192.
Suppose that sampling rate y of the present invention is 3, big or small W1 * H1 of the big figure of filtering sampling unit 31 sampling back characteristics is calculated as: W1=W/y=384/3=128; H1=H/y=192/3=64.The big figure of characteristic after will sampling in filtering sampling unit 31 arranges in order and constitutes one-dimensional characteristic vector β.
X '=xcos θ+ysin θ wherein, y '=-xsin θ+ycos θ.So template computing unit 32 needs only yardstick λ and the direction θ that calculates target image convolution template, definite this template that just can be unique.In the present invention, template computing unit 32 bases
Calculate yardstick λ and direction θ, k=1 wherein, 2,3.......N, W are that the big figure of characteristic is wide, and w is that target image is wide, and h is the target image height, and y is a sampling rate, β
kBe the eigenwert of utilizing the Adaboost algorithm to select.
Coordinate Calculation unit 33, be used for calculating target image unique point coordinate (x, y); Wherein
Wherein W1 is wide for the big figure of characteristic after sampling, and W is that the big figure of characteristic is wide, and w is that target image is wide, and h is that target image is high, and y is a sampling rate, β
kBe the eigenwert of utilizing the Adaboost algorithm to select.
Need to prove that it is few that smiling face's recognition device provided by the invention takies resource, fast operation is particularly suitable on mobile phone, using, because the mobile phone embedded platform of resource-constrained just behind this smiling face's recognition device of packing into, can be discerned the smiling face fast.Smiling face's recognition device that embodiment of the present invention embodiment provides; To the deficiency that exists in existing smiling face's recognition technology; A kind of new smiling face's recognition technology based on FL_Gabor algorithm and Adaboost is proposed; Improved the speed of feature extraction, reduced memory space, made on its embedded platform that is adapted at resource-constrained and use.
One of ordinary skill in the art will appreciate that all or part of flow process that realizes in the foregoing description method; Be to instruct relevant hardware to accomplish through computer program; Described program can be stored in the computer read/write memory medium; This program can comprise the flow process like the embodiment of above-mentioned each side method when carrying out.Wherein, described storage medium can be magnetic disc, CD, read-only storage memory body (Read-Only Memory, ROM) or at random store memory body (Random Access Memory, RAM) etc.
Above disclosedly be merely a kind of preferred embodiment of the present invention, can not limit the present invention's interest field certainly with this, the equivalent variations of therefore doing according to claim of the present invention still belongs to the scope that the present invention is contained.
Claims (10)
1. smiling face's recognition methods is characterized in that, comprising:
Smiling face's discrimination according to setting utilizes the Adaboost algorithm to select eigenwert;
Target image is carried out pre-service; Said target image judges for waiting whether it is smiling face's image; Said pre-service comprises that geometrical normalization is handled and unitary of illumination is handled;
According to said eigenwert, calculate the convolution template and the local unique point of said pretreated target image;
Carry out convolution algorithm according to said local feature point and convolution template, obtain proper vector;
Said proper vector is delivered to sorter mate, judge whether said target image is smiling face's image.
2. smiling face's recognition methods as claimed in claim 1 is characterized in that, said smiling face's discrimination according to setting utilizes the Adaboost algorithm to select before the eigenwert, also comprises training classifier; Said training classifier comprises:
Make up the image pattern storehouse, comprise smiling face's image and non-smiling face's image in the said image pattern storehouse;
According to the Gabor algorithm all images in the image pattern storehouse is carried out filtering, extract eigenwert;
Utilize the Adaboost algorithm to select characteristic, make up two-stage classification device at least.
3. smiling face's recognition methods as claimed in claim 3; It is characterized in that the convolution template of said target image is
X '=xcos θ+ysin θ wherein, y '=-xsin θ+ycos θ.
4. smiling face's recognition methods as claimed in claim 3 is characterized in that, the convolution template of the said target image of said calculating comprises:
With bank of filters said target image is carried out the Gabor feature ordering, the big figure of generating feature;
Sampling rate y according to setting samples to the big figure of said characteristic, and arranges formation one-dimensional characteristic vector β in order;
Calculate the yardstick λ and the direction θ of said convolution template, wherein
K=1,2,3.......N, W are that the big figure of characteristic is wide, and w is that target image is wide, and h is that target image is high, and y is a sampling rate, β
kBe the eigenwert of utilizing the Adaboost algorithm to select.
5. smiling face's recognition methods as claimed in claim 4 is characterized in that, the local feature point that calculates said target image comprises:
Unique point coordinate in the calculating target image (x, y), wherein,
x
k=((β
k%W
1The %w of) * y),
Wherein W1 is wide for the big figure of characteristic after sampling, and W is that the big figure of characteristic is wide, and w is that target image is wide, and h is that target image is high, and y is a sampling rate, β
kBe the eigenwert of utilizing the Adaboost algorithm to select.
6. smiling face's recognition device is characterized in that, comprising:
Eigenwert is selected module, is used for utilizing the Adaboost algorithm to select eigenwert according to smiling face's discrimination of setting;
Pre-processing module is used for target image is carried out pre-service, and said target image judges for waiting whether it is smiling face's image; Said pre-service comprises that geometrical normalization is handled and unitary of illumination is handled;
The calculation process module; Be used for selecting the selected eigenwert of module according to said eigenwert; Calculate the convolution template and the local unique point of the target image after said pre-processing module is handled, and carry out convolution algorithm, obtain proper vector according to said local feature point and convolution template;
Characteristic matching module is used for that the proper vector that said calculation process module obtains is delivered to sorter and matees, and judges whether said target image is smiling face's image.
7. smiling face's recognition device as claimed in claim 6; It is characterized in that said device also comprises: the training sort module is used to make up the image pattern storehouse; According to the Gabor algorithm all images in the image pattern storehouse is carried out filtering; Extract eigenwert, and utilize the Adaboost algorithm to select characteristic, make up two-stage classification device at least;
Comprise smiling face's image and non-smiling face's image in the said image pattern storehouse.
8. smiling face's recognition device as claimed in claim 7; It is characterized in that the convolution template of said target image is
X '=xcos θ+ysin θ wherein, y '=-xsin θ+ycos θ.
9. smiling face's recognition device as claimed in claim 8 is characterized in that, said calculation process module comprises:
The filtering sampling unit is used for bank of filters said target image being carried out the Gabor feature ordering, the big figure of generating feature; And the big figure of said characteristic is sampled, and arrange in order and constitute one-dimensional characteristic vector β according to the sampling rate y that sets;
The template computing unit is used to calculate the yardstick λ and the direction θ of said convolution template, wherein
K=1,2,3.......N, W are that the big figure of characteristic is wide, and w is that target image is wide, and h is that target image is high, and y is a sampling rate, β
kBe the eigenwert of utilizing the Adaboost algorithm to select;
The coordinate Calculation unit, be used for calculating target image unique point coordinate (x, y); X wherein
k=((β
k%W
1The %w of) * y),
Wherein W1 is wide for the big figure of characteristic after sampling, and W is that the big figure of characteristic is wide, and w is that target image is wide, and h is that target image is high, and y is a sampling rate, β
kBe the eigenwert of utilizing the Adaboost algorithm to select;
The convolution algorithm unit, the local feature point that convolution template that is used for said template computing unit is calculated and said coordinate Calculation unit calculate carries out convolution algorithm, obtains proper vector.
10. a portable terminal is characterized in that, said portable terminal comprises like each described smiling face's recognition device in the claim 6 to 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2010105519280A CN102467658A (en) | 2010-11-19 | 2010-11-19 | Identification method for smiling face and device thereof and mobile terminal thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2010105519280A CN102467658A (en) | 2010-11-19 | 2010-11-19 | Identification method for smiling face and device thereof and mobile terminal thereof |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN102467658A true CN102467658A (en) | 2012-05-23 |
Family
ID=46071282
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2010105519280A Pending CN102467658A (en) | 2010-11-19 | 2010-11-19 | Identification method for smiling face and device thereof and mobile terminal thereof |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102467658A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103886304A (en) * | 2014-04-03 | 2014-06-25 | 北京大学深圳研究生院 | True smile and fake smile identifying method based on space-time local descriptor |
| CN105956570A (en) * | 2016-05-11 | 2016-09-21 | 电子科技大学 | Lip characteristic and deep learning based smiling face recognition method |
| CN108205701A (en) * | 2016-12-20 | 2018-06-26 | 联发科技股份有限公司 | A kind of system and method for performing convolutional calculation |
| CN109740429A (en) * | 2017-11-30 | 2019-05-10 | 沈阳工业大学 | Smiling face's recognition methods based on corners of the mouth coordinate mean variation |
| US11281895B2 (en) * | 2019-07-11 | 2022-03-22 | Boe Technology Group Co., Ltd. | Expression recognition method, computer device, and computer-readable storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101075291A (en) * | 2006-05-18 | 2007-11-21 | 中国科学院自动化研究所 | Efficient promoting exercising method for discriminating human face |
| CN101339607A (en) * | 2008-08-15 | 2009-01-07 | 北京中星微电子有限公司 | Human face recognition method and system, human face recognition model training method and system |
-
2010
- 2010-11-19 CN CN2010105519280A patent/CN102467658A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101075291A (en) * | 2006-05-18 | 2007-11-21 | 中国科学院自动化研究所 | Efficient promoting exercising method for discriminating human face |
| CN101339607A (en) * | 2008-08-15 | 2009-01-07 | 北京中星微电子有限公司 | Human face recognition method and system, human face recognition model training method and system |
Non-Patent Citations (2)
| Title |
|---|
| YANG BAI ETAL: "<Image Processing(ICIP),2009 16th IEEE International Conference on>", 10 November 2009 * |
| ZHENGYOU ZHANG ET AL: "<Automatic Face and Gesture Recognition,1998.Proceedings.Third IEEE International Conference on>", 16 April 1998 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103886304A (en) * | 2014-04-03 | 2014-06-25 | 北京大学深圳研究生院 | True smile and fake smile identifying method based on space-time local descriptor |
| CN103886304B (en) * | 2014-04-03 | 2017-03-29 | 北京大学深圳研究生院 | It is a kind of that recognition methodss are laughed at based on the true and false of space-time local description |
| CN105956570A (en) * | 2016-05-11 | 2016-09-21 | 电子科技大学 | Lip characteristic and deep learning based smiling face recognition method |
| CN105956570B (en) * | 2016-05-11 | 2019-08-20 | 电子科技大学 | Smiling face's recognition methods based on lip feature and deep learning |
| CN108205701A (en) * | 2016-12-20 | 2018-06-26 | 联发科技股份有限公司 | A kind of system and method for performing convolutional calculation |
| CN108205701B (en) * | 2016-12-20 | 2021-12-28 | 联发科技股份有限公司 | System and method for executing convolution calculation |
| CN109740429A (en) * | 2017-11-30 | 2019-05-10 | 沈阳工业大学 | Smiling face's recognition methods based on corners of the mouth coordinate mean variation |
| US11281895B2 (en) * | 2019-07-11 | 2022-03-22 | Boe Technology Group Co., Ltd. | Expression recognition method, computer device, and computer-readable storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Raja et al. | Color object detection based image retrieval using ROI segmentation with multi-feature method | |
| CN109117803B (en) | Face image clustering method and device, server and storage medium | |
| CN110647817B (en) | Real-time face detection method based on MobileNet V3 | |
| WO2021248733A1 (en) | Live face detection system applying two-branch three-dimensional convolutional model, terminal and storage medium | |
| Varish et al. | Image retrieval scheme using quantized bins of color image components and adaptive tetrolet transform | |
| CN111476306A (en) | Object detection method, device, equipment and storage medium based on artificial intelligence | |
| CN109299315A (en) | Multimedia resource classification method, device, computer equipment and storage medium | |
| US20110085728A1 (en) | Detecting near duplicate images | |
| CN111553406A (en) | Target detection system, method and terminal based on improved YOLO-V3 | |
| CN102479320A (en) | Face recognition method and device as well as mobile terminal | |
| CN107679572B (en) | Image distinguishing method, storage device and mobile terminal | |
| CN107679448A (en) | Eyeball action-analysing method, device and storage medium | |
| WO2022089124A1 (en) | Certificate authenticity identification method and apparatus, computer-readable medium, and electronic device | |
| CN103617432A (en) | Method and device for recognizing scenes | |
| CN109063678B (en) | Face image recognition method, device and storage medium | |
| US20190205693A1 (en) | Scale-Invariant Feature Point Extraction in Edge Map | |
| CN102467658A (en) | Identification method for smiling face and device thereof and mobile terminal thereof | |
| Zheng et al. | Pedestrian detection based on gradient and texture feature integration | |
| Thabet et al. | Image processing on mobile devices: An overview | |
| CN113869282A (en) | Face recognition method, hyper-resolution model training method and related equipment | |
| Abdesselam | Improving local binary patterns techniques by using edge information | |
| CN102915449A (en) | Photo classification method | |
| Swamy et al. | Indian sign language interpreter with android implementation | |
| Cai et al. | Efficient facial expression recognition based on convolutional neural network | |
| CN112883956A (en) | Text character recognition method and device and server |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20120523 |