CN102201188A - Building television advertisement system oriented intelligent control device and method - Google Patents
Building television advertisement system oriented intelligent control device and method Download PDFInfo
- Publication number
- CN102201188A CN102201188A CN201110136757XA CN201110136757A CN102201188A CN 102201188 A CN102201188 A CN 102201188A CN 201110136757X A CN201110136757X A CN 201110136757XA CN 201110136757 A CN201110136757 A CN 201110136757A CN 102201188 A CN102201188 A CN 102201188A
- Authority
- CN
- China
- Prior art keywords
- image
- age
- neural network
- gender
- network chip
- 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 Processing (AREA)
- Image Analysis (AREA)
Abstract
本发明一种面向楼宇电视广告系统的智能控制装置和方法，主要包括图像采集装置、DSP图像处理单元、性别估计神经网络芯片、男性年龄估计神经网络芯片、女性年龄估计神经网络芯片和控制单元；利用神经网络芯片进行对当前受众人群进行性别年龄识别，从而选择已存有的针对该受众人群性别年龄层的广告送至楼宇电视进行播放；从而能够有效地进行针对性别、针对年龄层次受众人群的广告，获得最佳的广告效应。
The present invention is an intelligent control device and method for a building TV advertising system, mainly comprising an image acquisition device, a DSP image processing unit, a gender estimation neural network chip, a male age estimation neural network chip, a female age estimation neural network chip and a control unit; The neural network chip is used to identify the gender and age of the current audience, so that the existing advertisements targeting the gender and age group of the audience are selected and sent to the building TV for playback; thus, it is possible to effectively conduct gender-specific and age-specific advertisements for the audience Advertising, get the best advertising effect.
Description
Technical field
The present invention relates to a kind of intelligence controlling device and method towards building television advertising system.
Background technology
Increasingly mature along with this emerging medium of building TV, the clear location of its lively form of expression, Focus, compulsory view reception effect enjoy middle and high end advertiser's favor, become China's successful story in new media market for over ten years.The LCD TV radio hookup net of tens thousand of business premises of whole nation dozens of cities has been present among the life of city white collar truly, affects their Brang Awareness subtlely, is provoking their desire to purchase for extensive stock.If before 2004 when the building TV exposes as brand-new media vehicles, also have many people may hold the suspection attitude to its market outlook and profit, advancing by leaps and bounds of international funds that the today after 3 years, the building TV was raised and advertising income brought unprecedented impact all for whole medium market.
Building advertisement is a kind of medium that Focus Media establishes the earliest, just is mounted in the advertisement that a kind of building LCD TV in places such as high-grade office building, hotel, hotel, office building, megastore is play.Its success can reduce the following aspects.At first, compare traditional outdoor advertising, it is a kind of outdoor advertising of televised, it is the advertisement of audio ﹠ video combination, it has more expressive force and moves power than conventional outdoor advertising, its value is to improve brand recognition, more may change audient crowd's Brang Awareness with the powerful expressive force of its video display, provokes the desire for consumer goods of people to promotion item.On the other hand, compare with conventional advertisement, it is again a kind of advertisement of Focusization, and the business premises network advertising that Focus is made can accurately lock the stratum that corporate boss, manager and white collar etc. have more consumptive power, accurately hit audient crowd.In addition, the business premises network advertising of Focus also has high mandatory view reception effect.Compare outdoor advertising, the LCD of Focus is placed on mandatory rating district, and only is unique channel in business premises, and the advertisement of LCD bottom news rolling in real time simultaneously and excellent fashion interts mutually merges, and very high ad attention is arranged.From another angle, the building television hookup advertisement of Focus not only has the mandatory of vision, have more the mandatory of psychology, similarly be that audient crowd can unconsciously remove to see the flight magazine aboard, because audient crowd is in the area of an information vacuum, they can feel bored, barren and uncomfortable, so the audient crowd who is detained at the business premises lift port also has same situation, helpless seeking freed, boring sensation seeking, barren seek interesting, as long as a dot information is arranged, just can activate and note and interest, even advertisement.Because they are in a time and a space than advertisement more bored, this is so-called " waiting power economy " maximum feature.
And in the present building advertisement, though the lively form of expression and compulsory view reception effect that it has can form good communication effect, but the clear polarization of its Focus but still is worth further improving, show more intellectuality and hommization, the reception and registration effect of reinforcement advertising.Here show mainly that correspondent advertisement is observed and made in audient crowd's active plays the adjustment aspect, need utilize new technology that audient crowd is carried out the more layering of human nature and also play respective advertisement targetedly, thereby can more effectively publicize widely, strengthen the automatic station-keeping ability of Focus of building advertisement.
Summary of the invention
The invention provides a kind of speciality that can real-time judge commercial audience crowd, and choose the intelligence controlling device and the method towards building television advertising system of corresponding ad playing content according to this speciality.
In order to reach above-mentioned purpose, a kind of intelligence control method towards building television advertising system of the present invention mainly comprises training and discerns two stages, wherein:
This training stage comprises the branch of no sequencing and fully independently estimation of Age training and sex recognition training, the weighting parameter of each training is kept in the register of corresponding neural network chip automatically, and this neural network chip obtains weighting parameter automatically when by the time discerning from register;
The training of described estimation of Age is divided into men and women's two parts by sex to be carried out separately, two groups of files of the age label of facial image correspondence in sex face database of being made up of facial image by the data-interface input earlier by sex before the training and the face database; In the DSP graphics processing unit, sample is finished geometrical normalization and the standardized image pre-service of gray balance light that comprises facial image rotation, convergent-divergent then; Then pretreated imagery exploitation NNSC algorithm is calculated image feature value in the DPS graphics processing unit, and this image feature value imported into corresponding estimation of Age neural network chip together with class label information through data bus, estimate that through the age neural network chip training draws hidden layer and also is stored to automatically in the corresponding register with the weighting parameter of output layer;
Described sex recognition training, first two groups of files of the sex label of facial image correspondence in face database of being made up of facial image by the data-interface input and the face database are finished the geometrical normalization that comprises facial image rotation, convergent-divergent and the standardized image pre-service of light of use gray balance to sample then in the DSP graphics processing unit; Then pretreated imagery exploitation PCA algorithm computation is published picture as eigenwert, and image feature value and classification information imported in the corresponding sex identification neural network chip through data bus, obtain hidden layer through the training of sex identification neural network chip and also be stored to automatically in the corresponding register with the weighting parameter of output layer;
This cognitive phase specifically may further comprise the steps:
Step 1, earlier obtain by image collecting device and be subjected to behind building TV everybody group image and be sent to the DSP graphics processing unit to carry out the image pre-service;
Step 2, end user's face detection tech obtains the facial image of all audient crowd's individualities in audient crowd's image in the DSP graphics processing unit, detected every facial image is carried out eyes coordinate setting, and utilize the binocular information that obtains to carry out the image geometry normalized, use the gray balance technology to carry out the standardized image pre-service of light simultaneously, and it is pulled into capable vector form, then this sample vector is done projection to prior off-line training good PCA and two basis matrixs of NNSC respectively, obtain the two group face characteristic values relevant with sex and age;
Step 3, lineup's face eigenwert relevant with sex in the step 2 is exported to sex to be estimated in the neural network chip, neural network chip reads the weights data that trained and carries out sex identification from its register, can obtain the sex label information of current facial image, according to the sex label information, exporting lineup's face eigenwert relevant with the age in the step 2 to men age respectively according to sex estimates in neural network chip or the female age estimation neural network chip, this neural network chip reads the weighting parameter that has trained and carries out age identification from its corresponding register, and the estimation of Age value of this facial image is temporarily stored in its counter tank; This step of circular flow is sent to control module with the pairing sex of each facial image, estimation of Age value in the counter tank after face images is finished identification;
Step 4, control module are judged the sex age level that current audient crowd occupies the majority according to these data, and the advertisement of selecting thus to have had at this audient crowd's sex age level is delivered to the building TV and play.
Described NNSC arthmetic statement is as follows:
Step 1, iterations initial value k=1 is set, PRP conjugate gradient algorithm precision precision ζ＞0.01 and image reconstruction error ζ≤0.02; Select base vector a
i(k) and coefficient vector s
i(k) non-negative initial value, and to base vector a
i(k) and coefficient vector s
i(k) carry out the normalization operation;
Step 2, iterative process:
Fixing current base vector a
i(k), adopt the PRP algorithm to realize coefficient vector s
i(k) iteration; Make coefficient component s
i(k) all negative elements in are zero, and normalization
Order then
The currency s that fixedly obtains
i(k+1), adopt the PRP algorithm to realize base vector a
i(k) iteration; Make base vector a
i(k) all negative elements in are zero, make a
i(k+1) :=a
i(k); Make k:=k+1; If algorithm convergence then finish the conjugate gradient iterative process, otherwise restart the conjugate gradient iterative process;
Step 3, to basis function A and matrix of coefficients S after upgrading, judge
Whether set up, if set up, learning process, the basis function A that obtains after the algorithm convergence is the NNSC basis matrix of being asked if finishing; Otherwise the conjugate gradient iterative process of repeating step 2.
A kind of intelligence controlling device towards building television advertising system of the present invention mainly comprises image collecting device, DSP graphics processing unit, sex estimation neural network chip, men age estimation neural network chip, female age estimation neural network chip and control module;
Image collecting device wherein: be used to gather building TV audient crowd image, and be sent to the DSP graphics processing unit and carry out the image pre-service;
The DSP graphics processing unit: end user's face detection tech obtains the facial image of all audient crowd's individualities from audient crowd's image that image collecting device transmits, detected every facial image is carried out eyes coordinate setting, and utilize the binocular information that obtains to carry out the image geometry normalized, use the gray balance technology to carry out the standardized image pre-service of light simultaneously, and it is pulled into capable vector form, then this row vector is done projection to prior off-line training good PCA and two basis matrixs of NNSC respectively, obtain two groups of relevant with sex and age respectively face characteristic values;
Sex estimation neural network chip: the face characteristic value relevant with sex that pre-service obtains to process DSP graphics processing unit image carried out sex identification, and exports in men age estimation neural network chip or the female age estimation neural network chip according to sex recognition result lineup's face eigenwert that the DSP Flame Image Process Dan Zhongyu age is relevant;
Men age is estimated neural network chip or female age estimation neural network chip: lineup's face eigenwert relevant with the age in the DSP graphics processing unit is carried out age identification, and sex, the estimation of Age value of each facial image correspondence is sent to control module;
Control module: judge the sex age level that current audient crowd occupies the majority according to these data, the advertisement of selecting thus to have had at this audient crowd's sex age level is delivered to the building TV and is play.
The present invention utilizes biometrics identification technology to carry out layering to the commercial audience crowd, and on existing recognition technology, estimation of Age and sex identification can form maximum commercial value to commercial audience crowd layering.Can distinguish audient crowd's sex by sex identification, obtain current audient crowd's most of same sex character, utilize simultaneously estimation of Age to obtain overall ages again, thereby can carry out effectively obtaining best demonstration effect at sex, at ages audient crowd's advertisement.
Description of drawings
Fig. 1 is the workflow synoptic diagram of a kind of intelligence control method towards building television advertising system among the present invention;
Fig. 2 is the structural representation of a kind of intelligence controlling device towards building television advertising system among the present invention;
The NNSC neural network model synoptic diagram of Fig. 3 for feeding back among the present invention.
The invention will be further described below in conjunction with specific embodiment.
Embodiment
A kind of intelligence control method towards building television advertising system of the present invention mainly comprises training and discerns two stages, wherein:
This training stage comprises the branch of no sequencing and fully independently estimation of Age training and sex recognition training, the present invention is provided with three radial base neural net chips that are respectively applied for sex identification, male sex's facial image estimation of Age and women's facial image estimation of Age, the weighting parameter of each training is kept in the register of neural network chip automatically, and this neural network chip obtains weighting parameter automatically when by the time discerning from register;
The present invention is divided into men and women's two parts by sex and carries out the estimation of Age training separately, the age label of facial image correspondence is (as 1 in sex face database of being made up of facial image by the data-interface input earlier by sex before the training and the face database, 2,30 etc.) two groups of files; In the DSP graphics processing unit sample being finished geometrical normalizations such as comprising facial image rotation, convergent-divergent then handles, uses simultaneously gray balance etc. to carry out the image pre-service of light standardization; Then pretreated imagery exploitation NNSC algorithm is calculated image feature value in the DPS graphics processing unit, and this image feature value imported into corresponding estimation of Age neural network chip (determining to select which kind of neural network chip by class label information) together with class label information through data bus, estimate that through the age neural network chip training draws hidden layer and also is stored to (weights of input layer and hidden layer do not need to calculate, and are fixed as 1) in the corresponding register automatically with the weighting parameter of output layer.
During the sex recognition training, earlier the sex label of facial image correspondence (is represented the male sex as 0 in face database of being made up of facial image by the data-interface input and the face database, 1 expression women) two groups of files are finished geometrical normalizations such as comprising facial image rotation, convergent-divergent processing, are used gray balance etc. to carry out the image pre-service of light standardization simultaneously sample in the DSP graphics processing unit then; Then pretreated imagery exploitation PCA algorithm computation is published picture as eigenwert, and image feature value and classification information imported in the corresponding sex identification neural network chip through data bus, obtain hidden layer through the training of sex identification neural network chip and also be stored to automatically in the corresponding register with the weighting parameter of output layer.
As shown in Figure 1, 2, this cognitive phase specifically may further comprise the steps:
Step 1, earlier obtain by image collecting device 1 and be subjected to behind building TV everybody group image and be sent to DSP graphics processing unit 2 to carry out the image pre-service;
Step 2, end user's face detection tech obtains the facial image of all audient crowd's individualities in audient crowd's image in DSP graphics processing unit 2, detected every facial image is carried out eyes coordinate setting, and utilize the binocular information that obtains to carry out the image geometry normalized, use the gray balance technology to carry out the standardized image pre-service of light simultaneously, and it is pulled into capable vector form, then this sample vector is done projection (being that the capable vector of sample multiplies each other with basis matrix) to prior off-line training good PCA and two basis matrixs of NNSC respectively, obtain the two group face characteristic values relevant with sex and age;
Step 3, lineup's face eigenwert relevant with sex in the step 2 is exported to sex to be estimated in the neural network chip 3, sex estimates that neural network chip 3 reads the weights data that trained and carries out sex identification from its register, can obtain the sex label information of current facial image, according to the sex label information, exporting lineup's face eigenwert relevant with the age in the step 2 to men age respectively according to sex estimates in neural network chip 41 or the female age estimation neural network chip 42, this neural network chip 41 or 42 reads the weighting parameter that has trained and carries out age identification from its corresponding register, and the estimation of Age value of this facial image is temporarily stored in its counter tank; This step of circular flow is sent to control module 5 with the pairing sex of each facial image, estimation of Age value in the counter tank after face images is finished identification;
Step 4, control module 5 are judged the sex age level that current audient crowd occupies the majority according to these data, and the advertisement of selecting thus to have had at this audient crowd's sex age level is delivered to the building TV and play.
As shown in Figure 2, a kind of intelligence controlling device towards building television advertising system of the present invention mainly comprises image collecting device 1, DSP graphics processing unit 2, sex estimation neural network chip 3, men age estimation neural network chip 41, female age estimation neural network chip 42 and control module 5:
Wherein image collecting device 1: be used to gather building TV audient crowd image, and be sent to DSP graphics processing unit 2 and carry out the image pre-service;
DSP graphics processing unit 2: end user's face detection tech obtains the facial image of all audient crowd's individualities from audient crowd's image that image collecting device 1 transmits, detected every facial image is carried out eyes coordinate setting, and utilize the binocular information that obtains to carry out the image geometry normalized, use the gray balance technology to carry out the standardized image pre-service of light simultaneously, and it is pulled into capable vector form, then this row vector is done projection to prior off-line training good PCA and two basis matrixs of NNSC respectively, obtain two groups of relevant with sex and age respectively face characteristic values;
Sex estimation neural network chip 3: the face characteristic value relevant with sex that the 2 image pre-service through the DSP graphics processing unit obtain carried out sex identification, and lineup's face eigenwert relevant with the age in the DSP graphics processing unit 2 is exported in men age estimation neural network chip 41 or the female age estimation neural network chip 42 according to the sex recognition result;
Men age is estimated neural network chip 41 or female age estimation neural network chip 42: lineup's face eigenwert relevant with the age in the DSP graphics processing unit 2 is carried out age identification, and sex, the estimation of Age value of each facial image correspondence is sent to control module 5;
Control module 5: judge the sex age level that current audient crowd occupies the majority according to these data, the advertisement of selecting thus to have had at this audient crowd's sex age level is delivered to the building TV and is play.
The present invention in normal operation, being used for sex identification neural network chip 3 in three neural network chips at first works, its output result is 0 (male sex) or 1 (women), use simple gate circuit output port as a result and men age are estimated that neural network chip 41 and female age estimate that neural network chip 42 is connected respectively, reach the purpose of automatic selection chip.
Because calculated amount is bigger, the calculating of NNSC basis matrix and PCA basis matrix should not realize in DSP graphics processing unit 2, so adopts the mode of calculated off-line, and will calculate the matrix value of getting well and be stored in the register of each neural network chip.
For given estimation of Age training sample, on PC, utilize the NNSC algorithm that training sample is carried out off-line training in advance and obtain basis matrix, then this basis matrix is passed through in the register of data-interface afferent nerve network chip.In DSP graphics processing unit 2, the row vector that will pull into through the facial image after pretreated multiplies each other with this basis matrix, can obtain the proper vector relevant with the age, these characteristic vector datas import corresponding neural network chip again into to carry out the training of neural network weight parameter or is directly used in identification.
For sex recognition training sample, on PC, training sample is carried out calculated off-line equally in advance and obtain basis matrix by the PCA algorithm,, then this basis matrix is passed through in the register of data-interface afferent nerve network chip.In DSP graphics processing unit 2, the row vector that facial image after pretreated is pulled into multiplies each other with this basis matrix, can obtain the proper vector relevant with sex, these characteristic vector datas import corresponding neural network chip into to carry out the training of neural network weight parameter or is directly used in identification.
If the user has the face database of oneself, want device training again, the present invention will provide the algorithm bag to go out above-mentioned two groups of basis matrixs for user's calculated off-line so, and the user can be imported the result into device by data-interface, and recognition training step afterwards can be operated according to device peripheral hardware button.Device peripheral hardware sequence of operations button of the present invention, as " training parameter setting ", " begin training ", " start working " etc., wherein " training parameter setting " button is down with " sex identification " and " estimation of Age ", " estimation of Age " has the sex selection button of " male sex " and " women ", can carry out sample training to the neural network chip of this device when needing.
Non-negative sparse coding algorithm of the present invention (NNSC) is a kind of new subspace method, and it is a kind of combination of NMF algorithm and SC algorithm.It has all applied non-negativity constraint to the pixel and the reconstructed coefficients of basic image, make that reconstructed image is to be formed by the non-stack combinations that subtracts of basic image, the notion that more meets " local formation is whole " among the human thinking, simultaneously and the change of age locality that influence shows to people's face match, have intuitively feasibility at the application power of estimation of Age.Its feature extraction ability to facial image has just in time met the requirement of the required feature extraction of estimation of Age, can be used as the feature extracting method of good estimation of Age.
Described non-negative sparse coding (NNSC) algorithm based on expansion is on the basis of non-negative sparse coding (NNSC) algorithm that Hoyer proposes, consider the feedback input of the main visual cortex V1 district CK body LGN floor of primary vision system, the hierarchical structure of analog vision system, having set up one has the NNSC neural network model that connects feedback, as shown in Figure 3.Wherein on behalf of extraneous natural perception information, the input layer unit be projected in the input signal of LGN layer behind retina image-forming; Feedforward connects the receptive field of representing V1 district simple cell; Output layer is represented the simple cell in V1 district, and the non-negative sparse coding of the corresponding natural perception information of its active state, output layer have self feed back and exist feedback to connect to input layer.Parameter symbol implication is described below:
Expression LGN is to the feedforward connecting path in V1 district; a
iExpression V1 district is to the feedback connecting path of LGN; f
i(s
i) expression neuron response from inhibition function; X represents to be mapped to by retina the data acquisition of LGN; X (k) remarked pixel coordinate; a
iThe i column vector of expression basis function A; s
iThe capable vector of i of expression matrix of coefficients S, e
iThe capable vector of i of expression error matrix E=X-AS.
The NNSC objective function of described expansion is defined as follows:
This objective function is formed by three, and first is image minimal reconstruction error term; Second is sparse penalty term; The 3rd is the feedback connection item of V1 district to LGN;
Being constrained to of this formula: X (x, y) 〉=0,
And || s
i||=1, wherein,
The pretreated natural image data of the actual expression of X, parameter lambda and γ represent positive constant, the sparse penalty f (s in second of the objective function
i) selection depend on coefficient component s
iThe sparse density p (s of priori
i).
In the training stage of expansion NNSC network, in order to embody of the influence of main visual cortex V1 district neuron to the neuronic feedback input of LGN floor, the descend optimized Algorithm of (Conjugate Gradient) of PRP (Polak-Ribiere-Polyak) conjugate gradient is adopted in the renewal of sparse coefficient component, the study of feature basis function is also adopted use the same method.
The NNSC arthmetic statement of described expansion is as follows:
Step 1, iterations initial value k=1 is set, PRP conjugate gradient algorithm precision precision ζ＞0.01 and image reconstruction error ζ≤0.02; Select base vector a
i(k) and coefficient vector s
i(k) non-negative initial value, and to base vector a
i(k) and coefficient vector s
i(k) carry out the normalization operation;
Step 2, iterative process:
Fixing current base vector a
i(k), adopt the PRP algorithm to realize coefficient vector s
i(k) iteration; Make coefficient component s
i(k) all negative elements in are zero, and normalization
Order then
The currency s that fixedly obtains
i(k+1), adopt the PRP algorithm to realize base vector a
i(k) iteration; Make base vector a
i(k) all negative elements in are zero, make a
i(k+1) :=a
i(k); Make k:=k+1; If algorithm convergence then finish the conjugate gradient iterative process, otherwise restart the conjugate gradient iterative process;
Step 3, to basis function A and matrix of coefficients S after upgrading, judge
Whether set up, if set up, learning process, the basis function A that obtains after the algorithm convergence is the NNSC basis matrix of being asked if finishing; Otherwise the conjugate gradient iterative process of repeating step 2.
Training sample (or test sample book) is pulled into back the multiplying each other with basis matrix of capable vector can obtain new facial image eigenwert, can be applicable to next step age identification step.
Calculating for the PCA basis matrix, the present invention adopts Artificial Neural Network to realize, when artificial neural network passes through iterative PCA, each input new samples only causes the slight adjustment of basis matrix, neural network only need be upgraded on original basis and get final product when new samples occurring, weights needn't all recomputate, and therefore have adaptivity.Training sample (or test sample book) is pulled into back the multiplying each other with this basis matrix of capable vector can obtain new facial image eigenwert, can be applicable to next step sex identification step.
Claims (3)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201110136757XA CN102201188A (en) | 2011-05-25 | 2011-05-25 | Building television advertisement system oriented intelligent control device and method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201110136757XA CN102201188A (en) | 2011-05-25 | 2011-05-25 | Building television advertisement system oriented intelligent control device and method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN102201188A true CN102201188A (en) | 2011-09-28 |
Family
ID=44661834
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201110136757XA Pending CN102201188A (en) | 2011-05-25 | 2011-05-25 | Building television advertisement system oriented intelligent control device and method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102201188A (en) |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102495839A (en) * | 2011-11-04 | 2012-06-13 | 同辉佳视（北京）信息技术股份有限公司 | Interactive presentation system based on information publishing and touch inquiring, and method for the same |
| CN102521622A (en) * | 2011-11-18 | 2012-06-27 | 常州蓝城信息科技有限公司 | Face detecting system based on advertisement putting |
| CN102542252A (en) * | 2011-11-18 | 2012-07-04 | 江西财经大学 | Intelligent advertisement delivery system |
| CN102592116A (en) * | 2011-12-27 | 2012-07-18 | Tcl集团股份有限公司 | Cloud computing application method, system and terminal equipment, and cloud computing platform |
| CN102968738A (en) * | 2012-12-06 | 2013-03-13 | 中国科学院半导体研究所 | Advertising system |
| CN104335245A (en) * | 2012-06-26 | 2015-02-04 | 英特尔公司 | Method and apparatus for measuring audience size for a digital sign |
| CN105809178A (en) * | 2014-12-31 | 2016-07-27 | 中国科学院深圳先进技术研究院 | Population analyzing method based on human face attribute and device |
| WO2017124646A1 (en) * | 2016-01-20 | 2017-07-27 | 北京中科寒武纪科技有限公司 | Artificial neural network calculating device and method for sparse connection |
| CN106991426A (en) * | 2016-09-23 | 2017-07-28 | 天津大学 | Remote sensing images sparse coding dictionary learning method based on DSP embedded |
| CN107567636A (en) * | 2015-09-17 | 2018-01-09 | 三星电子株式会社 | Display device and its control method and computer readable recording medium storing program for performing |
| CN108460334A (en) * | 2018-01-23 | 2018-08-28 | 北京易智能科技有限公司 | A kind of age forecasting system and method based on vocal print and facial image Fusion Features |
| CN108681928A (en) * | 2018-05-30 | 2018-10-19 | 中海云智慧（北京）物联网科技有限公司 | A kind of intelligent advertisement put-on method |
| CN109241339A (en) * | 2018-08-28 | 2019-01-18 | 三星电子（中国）研发中心 | A kind of music recommended method and device |
| CN109886095A (en) * | 2019-01-08 | 2019-06-14 | 浙江新再灵科技股份有限公司 | A kind of passenger's Attribute Recognition system and method for the light-duty convolutional neural networks of view-based access control model |
| CN110415023A (en) * | 2019-07-09 | 2019-11-05 | 日立楼宇技术（广州）有限公司 | Elevator advertisement recommendation method, device, equipment and storage medium |
| CN111079474A (en) * | 2018-10-19 | 2020-04-28 | 上海商汤智能科技有限公司 | Passenger state analysis method and device, vehicle, electronic device, and storage medium |
| US11175789B2 (en) | 2018-11-13 | 2021-11-16 | Samsung Electronics Co., Ltd. | Electronic apparatus and method for controlling the electronic apparatus thereof |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1428694A (en) * | 2001-12-29 | 2003-07-09 | 成都银晨网讯科技有限公司 | Embedded human face automatic detection equipment based on DSP and its method |
| CN101459806A (en) * | 2009-01-08 | 2009-06-17 | 北京中星微电子有限公司 | System and method for video playing |
| CN101950358A (en) * | 2010-09-30 | 2011-01-19 | 冠捷显示科技（厦门）有限公司 | Method for automatically estimating age and judging sex by intelligent television |
-
2011
- 2011-05-25 CN CN201110136757XA patent/CN102201188A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1428694A (en) * | 2001-12-29 | 2003-07-09 | 成都银晨网讯科技有限公司 | Embedded human face automatic detection equipment based on DSP and its method |
| CN101459806A (en) * | 2009-01-08 | 2009-06-17 | 北京中星微电子有限公司 | System and method for video playing |
| CN101950358A (en) * | 2010-09-30 | 2011-01-19 | 冠捷显示科技（厦门）有限公司 | Method for automatically estimating age and judging sex by intelligent television |
Non-Patent Citations (3)
| Title |
|---|
| 余庆，杜吉祥: "基于一种改进NMF算法的人脸年龄估计方法", 《中国图象图形学报》 * |
| 尚丽: "稀疏编码算法及其应用研究", 《中国优秀博硕士学位论文全文数据库(博士)信息科技辑》 * |
| 黄贤武,仲兴荣,王加俊: "基于WT/PCA 的自适应神经网络人脸识别方法", 《计算机应用》 * |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102495839A (en) * | 2011-11-04 | 2012-06-13 | 同辉佳视（北京）信息技术股份有限公司 | Interactive presentation system based on information publishing and touch inquiring, and method for the same |
| CN102521622A (en) * | 2011-11-18 | 2012-06-27 | 常州蓝城信息科技有限公司 | Face detecting system based on advertisement putting |
| CN102542252A (en) * | 2011-11-18 | 2012-07-04 | 江西财经大学 | Intelligent advertisement delivery system |
| CN102592116A (en) * | 2011-12-27 | 2012-07-18 | Tcl集团股份有限公司 | Cloud computing application method, system and terminal equipment, and cloud computing platform |
| CN104335245A (en) * | 2012-06-26 | 2015-02-04 | 英特尔公司 | Method and apparatus for measuring audience size for a digital sign |
| CN104335245B (en) * | 2012-06-26 | 2017-07-11 | 英特尔公司 | Method and device for measuring audient's scale of digital signage |
| CN102968738A (en) * | 2012-12-06 | 2013-03-13 | 中国科学院半导体研究所 | Advertising system |
| CN105809178A (en) * | 2014-12-31 | 2016-07-27 | 中国科学院深圳先进技术研究院 | Population analyzing method based on human face attribute and device |
| CN107567636A (en) * | 2015-09-17 | 2018-01-09 | 三星电子株式会社 | Display device and its control method and computer readable recording medium storing program for performing |
| WO2017124646A1 (en) * | 2016-01-20 | 2017-07-27 | 北京中科寒武纪科技有限公司 | Artificial neural network calculating device and method for sparse connection |
| CN106991426A (en) * | 2016-09-23 | 2017-07-28 | 天津大学 | Remote sensing images sparse coding dictionary learning method based on DSP embedded |
| CN106991426B (en) * | 2016-09-23 | 2020-06-12 | 天津大学 | Remote sensing image sparse coding dictionary learning method based on embedded DSP |
| CN108460334A (en) * | 2018-01-23 | 2018-08-28 | 北京易智能科技有限公司 | A kind of age forecasting system and method based on vocal print and facial image Fusion Features |
| CN108681928A (en) * | 2018-05-30 | 2018-10-19 | 中海云智慧（北京）物联网科技有限公司 | A kind of intelligent advertisement put-on method |
| CN109241339A (en) * | 2018-08-28 | 2019-01-18 | 三星电子（中国）研发中心 | A kind of music recommended method and device |
| CN111079474A (en) * | 2018-10-19 | 2020-04-28 | 上海商汤智能科技有限公司 | Passenger state analysis method and device, vehicle, electronic device, and storage medium |
| US11175789B2 (en) | 2018-11-13 | 2021-11-16 | Samsung Electronics Co., Ltd. | Electronic apparatus and method for controlling the electronic apparatus thereof |
| CN109886095A (en) * | 2019-01-08 | 2019-06-14 | 浙江新再灵科技股份有限公司 | A kind of passenger's Attribute Recognition system and method for the light-duty convolutional neural networks of view-based access control model |
| CN110415023A (en) * | 2019-07-09 | 2019-11-05 | 日立楼宇技术（广州）有限公司 | Elevator advertisement recommendation method, device, equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102201188A (en) | Building television advertisement system oriented intelligent control device and method | |
| CN112348640B (en) | An online shopping system and method based on facial emotional state analysis | |
| Desai et al. | ASL citizen: a community-sourced dataset for advancing isolated sign language recognition | |
| CN112508077B (en) | Social media emotion analysis method and system based on multi-modal feature fusion | |
| CN116484318B (en) | Lecture training feedback method, lecture training feedback device and storage medium | |
| CN117556802B (en) | User portrait method, device, equipment and medium based on large language model | |
| CN108877336A (en) | Teaching method, cloud service platform and tutoring system based on augmented reality | |
| Audhkhasi et al. | A globally-variant locally-constant model for fusion of labels from multiple diverse experts without using reference labels | |
| CN116704085B (en) | Avatar generation method, apparatus, electronic device, and storage medium | |
| CN109348275A (en) | Method for processing video frequency and device | |
| CN109885595A (en) | Course recommended method, device, equipment and storage medium based on artificial intelligence | |
| US12165633B2 (en) | Intelligent training and education bot | |
| CN115545982A (en) | System and method for heterogeneous grouping of online students | |
| CN118690002B (en) | A personalized test question recommendation method and system based on artificial intelligence | |
| CN118296183A (en) | Short video intelligent recommendation system and method based on interest modeling classification | |
| CN120580898A (en) | A multimodal English learning interactive system and vocabulary memory training method | |
| CN120123475A (en) | An AIGC-enabled Chinese and foreign cultural comparison auxiliary teaching system and teaching method | |
| CN116108158B (en) | Online interactive question answering text feature construction method and system | |
| CN118967224A (en) | An intelligent advertising production system | |
| CN116071794B (en) | Behavior data learning style portrait generation method, system, equipment and medium | |
| CN113869960A (en) | Poster generation method and device, storage medium and computer equipment | |
| EP4468694A1 (en) | Information processing device, information processing method, and program | |
| Jin | Cross-Modal Attention GAN for Text-to-Artistic Image Generation | |
| Affifatuholihah et al. | Relationship Between Social, Technology, Personality, and Psychology Factors on Intention in Pre-Employment Training | |
| CN121073560A (en) | Elevator content space multidimensional information display system and method |
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: 20110928 |