CN104954873B - A kind of smart television video method for customizing and system - Google Patents
A kind of smart television video method for customizing and system Download PDFInfo
- Publication number
- CN104954873B CN104954873B CN201410115778.7A CN201410115778A CN104954873B CN 104954873 B CN104954873 B CN 104954873B CN 201410115778 A CN201410115778 A CN 201410115778A CN 104954873 B CN104954873 B CN 104954873B
- Authority
- CN
- China
- Prior art keywords
- video
- information
- user
- particle
- video information
- 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.)
- Active
Links
Abstract
The present invention discloses a kind of smart television video method for customizing and system, the method includes：A, middle layer receives user video customized demand information, will be sent to cloud server after the information preservation；B, cloud server is according to the corresponding video information of the information search received, and returns it to middle layer；C, middle layer receives video information, nuclear space mapping and clustering are carried out to it using multinuclear population clustering algorithm, obtain include multiple category video information cluster result, the cluster result and video customized demand information are compared, most matched category video information is obtained；D, obtained category video information is shown.The large-scale data of high dimensional nonlinear can be effectively treated in video data Processing Algorithm used in present system, clustering precision can be significantly improved, reduce time complexity, its make smart television is more targeted to provide personalized service, make smart television more intelligence and hommization.
Description
Technical field
The present invention relates to technical field of information processing more particularly to a kind of smart television video method for customizing and systems.
Background technology
The fast-developing of internet has prodigious impact, internet to have following excellent the development of entire smart television industry
Gesture：It is more using the user of internet；The speed that information and transmission information are obtained by internet is fast；Internet may be implemented to surmount
The resource-sharing of space-time can complete real-time, interactive；Internet has the characteristics that personalized and hommization；Internet is for each
User is impartial to；Unnecessary expense can be saved by internet.In the influence of this brute force attack in internet
Under how to be promoted the direction for developing into smart television development of smart television by means of the advantage of internet, but at present by mutual
Networking makes the more personalized hommization of smart television, and there is also many deficiencies, nothings for the technology of realization real-time, interactive and resource-sharing
Method meets requirement of the user for experience satisfaction and autonomous control ability,
Therefore, the existing technology needs to be improved and developed.
Invention content
In view of above-mentioned deficiencies of the prior art, the purpose of the present invention is to provide a kind of smart television video method for customizing and
System, it is intended to solve the real-time, interactive resource encountered in current smart television evolution, hommization information processing, user's subjectivity
The problems such as dynamic role.
Technical scheme is as follows：
A kind of smart television video method for customizing, wherein the described method comprises the following steps：
A, middle layer receives user video customized demand information, will be sent to cloud server after the information preservation；
B, cloud server is according to the corresponding video information of user video customized demand information search received, and should
Video information is back to middle layer；
C, middle layer receives the video information returned, and it is empty to carry out core to the video information using multinuclear population clustering algorithm
Between map and clustering, obtain include multiple category video information cluster result, by the cluster result and the user
Video customized demand information is compared, and is obtained and the most matched category video information of the user video customized demand information；
D, obtained category video information is shown in application interface.
The smart television video method for customizing, wherein carry out obtained category video information in application interface
After display, category video information is sent to the cloud server and saves as the characteristic behavior collection of the user.
The smart television video method for customizing, wherein further include after the step D：
The suggestion feedback information that user is directed to shown category video information is received, multinuclear population clustering algorithm is utilized
Nuclear space mapping is carried out to the suggestion feedback information and clustering obtains result set, the result set and cloud server are preserved
The characteristic behavior collection of the user be compared, obtain the characteristic information of user, utilize the feature of this feature information update user
Behavior collection.
The smart television video method for customizing, wherein further include before the step A：
Middle layer receives the log-on message of user, and the log-on message is sent to cloud server；
Cloud server searches whether there is the user characteristics behavior collection being consistent with the log-on message, and if it exists, then by institute
The user characteristics behavior collection found is shown in application interface.
The smart television video method for customizing, wherein further include before the step A：
Middle layer obtains device id, and sends it to cloud server；
Cloud server searches whether there is the Family characteristics behavior collection being consistent with the device id, if there are Family characteristics rows
To collect and not finding the user characteristics behavior collection of active user, then by the Family characteristics behavior collection found using boundary
Face is shown.
The smart television video method for customizing, wherein the step C is specially：
C1, middle layer receive the video information returned, set up multinuclear nuclear matrix and are carried out to the video information using it
Nuclear space maps, and obtains nuclear space video information；
C2, obtained nuclear space video information is input to as particle in particle cluster algorithm, passes through iteration and adaptation
Optimal particle is found in degree evaluation；
C3, it is input to obtained optimal particle as initial cluster center in k-means clustering algorithms, more using it
New optimal particle, obtain include multiple category video information cluster result；
C4, the cluster result and the user video customized demand information are compared, is obtained and the user video
The most matched category video information of customized demand information.
The smart television video method for customizing, wherein multinuclear nuclear matrix is set up in the step C1 is specially：
Use expression formula forMultinuclear nuclear matrix, whereinFor multiple and different kernel functions,For
The linear combination coefficient of kernel function, when givenA different kernel function, then be corresponding withA different nuclear matrix, herein
In conjunction with the knowwhy of Semidefinite Programming, multinuclear nuclear matrix is solved using Semidefinite Programming interior point method, obtains the optimum linearity of nuclear matrix
Combination coefficient, will by optimum linearity combination coefficientA nuclear matrix is added, and obtains more preferably multinuclear nuclear matrix。
The smart television video method for customizing, wherein the step C2 is specially：
C21, obtained nuclear space video information is input to as particle in particle cluster algorithm, and is calculated by n times
The fitness of each particle obtains N number of primary group；
The fitness of C22, the current fitness by comparing each particle and its desired positions lived through updates
The desired positions of each particle pass through the adaptation for the desired positions that the current fitness and group that compare each particle are lived through
It spends to update the global desired positions of each particle；
C23, speed and the position that particle is adjusted according to particle cluster algorithm, calculate Colony fitness variance and whether judge it
Less than predetermined threshold value, if so, finding optimal particle and executing step C3
The smart television video method for customizing, wherein the step C3 is specially：
C31, using obtained optimal particle as the initial cluster center of k-means clustering algorithms, and according to closest
Rule determines the clustering of corresponding optimal particle；
C32, new cluster centre is calculated according to identified clustering, it is used in combination to replace initial cluster center；
C33, judge whether cluster centre is also changing or whether reaching maximum iteration, if cluster centre no longer becomes
Change or reach maximum iteration, thens follow the steps C34；
Whether C34, the fitness for judging new cluster centre, if so, updating optimal particle, are obtained due to previous particle
To the cluster result for including multiple category video information.
A kind of smart television video custom-built system, wherein the system comprises：
Middle layer for receiving user video customized demand information, and will be sent to cloud server after the information preservation；
And receive the video information that cloud server is returned according to user video customized demand information, it is clustered and is calculated using multinuclear population
Method carries out nuclear space mapping and clustering to the video information, obtain include multiple category video information cluster result,
The cluster result and the user video customized demand information are compared, obtained and the user video customized demand information
Most matched category video information；
Cloud server, for according to the corresponding video information of user video customized demand information search received, and
The video information is back to middle layer；
Application interface display module, for showing obtained category video information.
Advantageous effect：A kind of smart television video method for customizing of present invention offer and system, the system can realize user
It freely controls and real-time, interactive, there is good timeliness, and video data Processing Algorithm used in system can be effective
The large-scale data for handling high dimensional nonlinear, compared to traditional algorithm, it has the clustering precision that is significantly improved, and it is multiple to reduce the time
The effect of miscellaneous degree, the method for the present invention make smart television is more targeted to provide personalized service, make its more intelligence and
Hommization.Also the interconnection networking of smart television has been pushed to develop.
Description of the drawings
Fig. 1 is smart television video method for customizing flow chart in the specific embodiment of the invention.
Fig. 2 is the schematic diagram of smart television video custom-built system application interface in the specific embodiment of the invention.
Fig. 3 is the schematic diagram of smart television video custom-built system user's login interface in the specific embodiment of the invention.
Fig. 4 is the schematic diagram at smart television video custom-built system user's registration interface in the specific embodiment of the invention.
Fig. 5 is the specific method flow chart of step S300 in Fig. 1.
Fig. 6 is the specific method flow chart of step S320 in Fig. 5.
Fig. 7 is the specific method flow chart of step S330 in Fig. 5.
Fig. 8 is multinuclear population clustering algorithm flow chart in the specific embodiment of the invention.
Fig. 9 is the functional block diagram of smart television video custom-built system in the specific embodiment of the invention.
Specific implementation mode
The present invention provides a kind of smart television video method for customizing and system, for make the purpose of the present invention, technical solution and
Effect is clearer, clear, and the present invention is described in more detail below.It should be appreciated that specific embodiment described herein
It is only used to explain the present invention, be not intended to limit the present invention.
A kind of smart television video method for customizing as shown in Figure 1, wherein the described method comprises the following steps：
S100, middle layer receive user video customized demand information, will be sent to cloud server after the information preservation.
S200, cloud server according to the corresponding video information of user video customized demand information search received, and
The video information is back to middle layer.
Middle layer is also referred to as application service layer, specially takes charge of the realization of service logic, all videos are stored on cloud server
Information, for example, all video websites at present such as may include Tencent, Sohu, iqiyi.com, youku.com video information.One video corresponds to
Complete video information specifically may include：Source video sequence, video name, video format, video profile, video director, video
Cast, video playing address, video area, video type etc., video information data of the present invention are by one
Complete video information is converted into one group of data, this data is shown using coding, can finally be parsed coding to obtain this
One group of complete video information.
S300, middle layer receive the video information returned, are carried out to the video information using multinuclear population clustering algorithm
Nuclear space map and clustering, obtain include multiple category video information cluster result, by the cluster result with it is described
User video customized demand information is compared, and obtains believing with the most matched category video of the user video customized demand information
Breath.
In the present embodiment, the multinuclear population clustering algorithm is to have merged to calculate including multinuclear nuclear matrix algorithm, population
The general designation of algorithm including method and k-means clustering algorithms（Principle and this implementation of these three algorithms can be discussed in detail hereinafter
Example is directed to the concrete application after these three algorithms are organically blended）.It is noted that passing through multinuclear population clustering algorithm
The most like cluster result of many features is only obtained, is needed by being matched with user video customized demand information, ability
It determines and user video customized demand information is most like, most matched category video information, this classification video information is returned to
User.
S400, obtained category video information is shown in application interface.
Further, by obtained category video information after application interface is shown, by category video information
It is sent to the cloud server and saves as the characteristic behavior collection of the user.
Therefore, further include before the step S100：
Middle layer receives the log-on message of user, and the log-on message is sent to cloud server；
Cloud server searches whether there is the user characteristics behavior collection being consistent with the log-on message, and if it exists, then by institute
The user characteristics behavior collection found is shown in application interface.If cloud server does not find corresponding user characteristics row
For collection, then user enters after system, and system can show the high video information of some popular program request amounts to user automatically, if with
Family loses interest in these video informations, can enter customization video information entrance, that is, execute step S100.
Further include following before middle layer receives user video customized demand information in addition, after logging in system by user
Step：
Middle layer obtains device id, and sends it to cloud server；
Cloud server searches whether there is the Family characteristics behavior collection being consistent with the device id, if there are Family characteristics rows
To collect and not finding the user characteristics behavior collection of active user, then by the Family characteristics behavior collection found using boundary
Face is shown.
Therefore, in video custom-built system of the invention, which is corresponded to by user's registration ID, passes through intelligence
The device id of energy TV corresponds to the feature of the smart television use groups, that is, corresponds to family's behavioural characteristic collection, and the user is special
Sign behavior collection is the video preference information set for individual subscriber, and family's behavioural characteristic collection is to be directed to tie up with the device id
The video preference information set of all members of fixed family.
Specific application interface can be with as shown in Fig. 2, system application interface includes login button, customization video information entrance
Button and feedback opinion entrance enter login interface, as shown in Figure 3 by clicking login button.The user's needs used for the first time
It first registers, after the new user's registration for clicking log-in interface, into register interface, as shown in figure 4, the information of registration has：Equipment
ID, user name, password.Wherein device id directly can automatically be obtained by video device interface, obtain device id primarily to
Identify that user uses the characteristic of family.Automated log on returns to the first interface of application after registration is completed, at this time the login at first interface
Place show user information.System can record the login status of user after first logging into, and facilitate use after user.User
It is preceding to show the high video information of some popular program request amounts automatically to user, if user into application three times into after application
Lose interest in these video informations, can be required according to the customization of oneself, clicks customization video information entrance, regarded into customization
Frequency information interface inputs the video customized demand information of oneself, it is preferable that user video customized demand information is in addition to general general
Except logical video searching information, other demand informations of user are further comprised（Such as plot requirement, picture texture, Ren Wushe
It is fixed etc.）
The present embodiment is briefly to pass through first using the processing procedure to video information of multinuclear population clustering algorithm
Multiple and different kernel functions is carried out linear combination by the concept of kernel function and the feature of kernel function, and combination forms multinuclear nuclear matrix,
It solves to obtain the linear combination coefficient of multinuclear nuclear matrix using the knowledge of Semidefinite Programming, and the linear combination coefficient is substituted into and is combined
Multinuclear nuclear matrix in the multinuclear nuclear matrix of formation, and then after being solved.Core is carried out to data information using multinuclear nuclear matrix
Space reflection obtains the data set of nuclear space, is clustered, is obtained to nuclear space data set using multinuclear population clustering algorithm
Different category combinations.After obtaining different classifications, the corresponding classification of user is returned to according to the customization of user requirement and is regarded
Frequency information.Improve the satisfaction of the accuracy and user that return to video.
In preferred embodiment as shown in Figure 5, the step S300 is specially：
S310, middle layer receive return video information, set up multinuclear nuclear matrix and using its to the video information into
Row nuclear space maps, and obtains nuclear space video information.
The video information general data returned by cloud server is huge, and classification is complicated, these video informations have height
Dimension and nonlinear feature（I.e. data complexity is very high）.The multinuclear population clustering algorithm of the present invention is high using kernel function is commonly used
This kernel function, Polynomial kernel function and perceptron kernel function carry out nuclear matrix linear combination and obtain multinuclear nuclear matrix, because core is empty
Between can the data of high dimensional nonlinear be reduced to the dimension of data by mapping, while it being made to become linear separability, and each
The characteristics of kernel function has oneself map the data distribution that obtains later and be also different, and therefore, this nuclear matrix set is multiple
Kernel function advantage, linear combination becomes more excellent kernel function, and in specific embodiment, setting up multinuclear nuclear matrix is specially：
Use expression formula forMultinuclear nuclear matrix, whereinFor multiple and different kernel functions,For
The linear combination coefficient of kernel function, when givenA different kernel function, then be corresponding withA different nuclear matrix, herein
In conjunction with the knowwhy of Semidefinite Programming, multinuclear nuclear matrix is solved using Semidefinite Programming interior point method, obtains the optimum linearity of nuclear matrix
Combination coefficient, will by optimum linearity combination coefficientA nuclear matrix is added, and obtains more preferably multinuclear nuclear matrix。
Then nuclear space mapping is carried out to video information using multinuclear nuclear matrix obtained above, obtains nuclear space video letter
Breath.It in conjunction with the knowledge of theorem in Euclid space range formula, applies it on nuclear space, obtains nuclear space video set in nuclear space
Core distance.In short, realizing dimensionality reduction and video information data made to become linear separability by multinuclear nuclear matrix.
S320, obtained nuclear space video information is input to as particle in particle cluster algorithm, by iteration and is fitted
Optimal particle is found in response evaluation.
S330, it is input to obtained optimal particle as initial cluster center in k-means clustering algorithms, utilizes it
Update optimal particle, obtain include multiple category video information cluster result.
Particle cluster algorithm is a kind of evolvement method based on swarm intelligence.The potential solution of each of problem is search space
In particle, each particle has its corresponding speed, position and a fitness determined by object function, and algorithm passes through suitable
Response evaluates the quality of particle.Algorithm initializes a group random particles first, then finds optimal solution by iteration.Each
In secondary iteration, particle updates oneself by tracking two " extreme values "：One is optimal solution that particle itself is found, i.e., individual
Extreme value；The other is the optimal solution that entire population is found at present, referred to as global extremum.Particle is found above
After two extreme values, just according to formula（1）And formula（2）To update speed and the position of oneself：
（1）
（2）
Wherein,It is the speed of current particle,It is the current location of particle.,It is current
The dimension in space.,Shi [0 ,1]Between random number,WithFor Studying factors, usually take
。It is weighting coefficient, the value generally between 0.1 to 0.9.IfLinearly reduce with the progress of algorithm iteration, it will be notable
Improve convergence energy.IfFor maximum weighted coefficient,For minimum weight coefficient,For current iteration time
Number,For algorithm iteration total degree, then have,
（3）
Optimize particle cluster algorithm using k-means clustering algorithms, it is thus necessary to determine that the convergence opportunity of particle cluster algorithm, research hair
The state that population can be now tracked according to the overall variation of all particle fitness, judges whether algorithm restrains.
If the number of particles of population is,It isThe fitness of a particle,It is fitted for current average of population
Response, the Colony fitness variance of populationIt is defined as follows：
（4）
Fitness variance can reflect " convergence " degree of all particles in population.Smaller, then population tends to
Convergence；IfIt is zero, then group's fitness is almost the same, and particle swarm optimization algorithm is absorbed in Premature Convergence or reaches global receipts
It holds back；Conversely, fitness difference population is then in the random search stage.
The present embodiment is together handled data by using particle cluster algorithm and k-means clustering algorithms, k-means
The incision one of clustering algorithm is the Premature Convergence for avoiding particle cluster algorithm, and being also exactly particle cluster algorithm has Global treatment, office
The processing in portion needs introducing k-means clustering algorithms to be determined, and also particle cluster algorithm can determine k-means clustering algorithms
Initial center.Cause result inaccurate into convergence too early if particle cluster algorithm is used alone；K- is used alone
The selection of means clustering algorithm initial centers is not necessarily very accurate, causes result also inaccurate, therefore passes through two algorithms
In conjunction with the defect for making respective advantage evade other side.
S340, the cluster result and the user video customized demand information are compared, obtains regarding with the user
The most matched category video information of frequency customized demand information.
In preferred embodiment as shown in FIG. 6, the step S320 is specially：
S321, obtained nuclear space video information is input to as particle in particle cluster algorithm, and is calculated by n times
The fitness of each particle obtains N number of primary group.
The fitness of S322, the current fitness by comparing each particle and its desired positions lived through updates
The desired positions of each particle pass through the adaptation for the desired positions that the current fitness and group that compare each particle are lived through
It spends to update the global desired positions of each particle.
S323, speed and the position that particle is adjusted according to particle cluster algorithm, calculate Colony fitness variance and judge that it is
It is no to be less than predetermined threshold value, if so, finding optimal particle and executing step S300.
In preferred embodiment as shown in Figure 7, the step S330 is specially：
S331, using obtained optimal particle as the initial cluster center of k-means clustering algorithms, and according to closest
Rule determines the clustering of corresponding optimal particle；
S332, new cluster centre is calculated according to identified clustering, it is used in combination to replace initial cluster center；
S333, judge whether cluster centre is also changing or whether reaching maximum iteration, if cluster centre is no longer
Change or reach maximum iteration, thens follow the steps S334；
S334, whether the fitness of new cluster centre is judged due to previous particle, if so, update optimal particle,
Obtain include multiple category video information cluster result.
The multinuclear population clustering algorithm flow of the present embodiment as shown in Figure 8, its step are as follows：
S1, the video information that cloud server returns is received.
S2, component multinuclear nuclear matrix carry out dimension-reduction treatment to video information by nuclear space mapping, obtain nuclear space video
Information.
S3, nuclear space video information are input to as particle in particle cluster algorithm, and the adaptation of each particle is calculated by n times
Degree, obtains N number of primary group.It is first that certain is a kind of by each particle random division, as initial clustering, and calculates each
The cluster centre of class calculates the fitness of particle, while as the optimal position of individual of particle as the position encoded of primary
It sets, and the speed of random initializtion particle.It is repeatedIt is secondary, symbiosis atA primary group.
The fitness of S4, the fitness for comparing particle and desired positions that it is lived through, the desired positions of more new particle.
The fitness of S5, the fitness for comparing particle and desired positions that it is lived through, the best position of the overall situation of more new particle
It sets.
S6, the position and speed that particle is adjusted according to particle cluster algorithm.
Whether S7, group's fitness method are less than defined threshold, if so, S8 is thened follow the steps, if it is not, then return to step
S6。
According to above-mentioned formula（4）Judge whether current particle group reaches convergence state, if Colony fitness varianceIt is small
In defined threshold, then particle is further processed.
S8, optimal particle is obtained, is optimized using k-means clustering algorithms.SelectionA optimal particle carries out K-
Means local search algorithms avoid enter into Premature Convergence trap.
S9, using optimal particle as initial cluster center the cluster stroke of the corresponding particle is determined according to arest neighbors rule
Point.
S10, new cluster centre is calculated according to clustering, replace initial cluster center.Due to K-means clustering algorithms
With stronger local search ability, therefore the convergence rate for introducing the particle cluster algorithm after the optimization of K-means clustering algorithms can
To greatly improve.
S11, whether reach termination condition（Whether cluster centre no longer changes or whether reaches maximum iteration）If
It is to then follow the steps S12, if it is not, then executing return to step S10.
The fitness of S12, the new cluster centre of comparison.
It is S13, whether optimal.Judge whether the fitness of new cluster centre is optimization, if so, S14 is thened follow the steps,
If it is not, then return to step S10.Judge the fitness of new cluster centre, if more excellent than previous particle, more new particle is simultaneously held
Row step S14 terminates algorithm.
S14, algorithm terminate, and obtain optimal particle.
In preferred embodiment, further include after the step S400：
The suggestion feedback information that user is directed to shown category video information is received, multinuclear population clustering algorithm is utilized
Nuclear space mapping is carried out to the suggestion feedback information and clustering obtains result set, the result set and cloud server are preserved
The characteristic behavior collection of the user be compared, obtain the characteristic information of user, utilize the feature of this feature information update user
Behavior collection.And the characteristic behavior collection of updated user is saved in cloud server, so as to its as cloud server again
Preliminary reference when the secondary return video information for the user, keeps video customization flow more and more perfect.
Specifically, user obtains the video information of server end return, field feedback entrance is clicked after browsing, into
Enter feedback opinion interface, it can be by personal experience's opinion（Such as the satisfaction to returning to video information, description of contents, needle need to be improved
Demand point each to returned data to oneself scores, scoring reason etc.）Middle layer is fed back to by suggestion feedback column,
Middle layer is handled the suggestion feedback information also with multinuclear population clustering algorithm, and utilizes obtained result set
The characteristic behavior collection for updating user, to realize the real-time, interactive of user and system.By the feedback opinion for handling user in real time
Information, and analyzing processing is carried out for feedback opinion information, summarize the characteristic behavior of user and the global feature row of subscriber household
For that system can be made more targetedly to provide Video service to the user.
A kind of smart television video custom-built system as shown in Figure 9, wherein the system comprises：
Middle layer 100 for receiving user video customized demand information, and will be sent to cloud service after the information preservation
Device；And receive the video information that cloud server is returned according to user video customized demand information, utilize multinuclear particle clustering
Class algorithm carries out nuclear space mapping and clustering to the video information, obtain include multiple category video information cluster knot
Fruit compares the cluster result and the user video customized demand information, obtains and the user video customized demand
The most matched category video information of information, specifically as described in step S100 and S300.
Cloud server 200, for according to the corresponding video information of user video customized demand information search received,
And the video information is back to middle layer, specifically as described in step S200.
Application interface display module 300, for showing obtained category video information, specifically as described in step S400.
The present embodiment provides a kind of smart television video method for customizing and system, which can realize that user freely controls
And real-time, interactive, there is good timeliness, and higher-dimension can be effectively treated in video data Processing Algorithm used in system
Nonlinear large-scale data, compared to traditional algorithm, it has the clustering precision that is significantly improved, and reduces the effect of time complexity
Fruit, method of the invention make smart television is more targeted to provide personalized service, and make its more intelligence and hommization.
The interconnection networking of smart television has been pushed to develop.
It should be understood that the application of the present invention is not limited to the above for those of ordinary skills can
With improvement or transformation based on the above description, all these modifications and variations should all belong to the guarantor of appended claims of the present invention
Protect range.
Claims (10)
1. a kind of smart television video method for customizing, which is characterized in that the described method comprises the following steps：
A, middle layer receives user video customized demand information, will be sent to cloud server after the information preservation；
B, cloud server is according to the corresponding video information of user video customized demand information search received, and by the video
Information is back to middle layer；
C, middle layer receives the video information returned, and carrying out nuclear space to the video information using multinuclear population clustering algorithm reflects
Penetrate and clustering, obtain include multiple category video information cluster result, by the cluster result and the user video
Customized demand information is compared, and is obtained and the most matched category video information of the user video customized demand information；It is described
Multinuclear population clustering algorithm has merged multinuclear nuclear matrix algorithm, particle cluster algorithm and k-means clustering algorithms；
D, it is carried out obtained in application interface with the most matched category video information of the user video customized demand information
Display.
2. smart television video method for customizing according to claim 1, which is characterized in that believe obtained category video
Category video information is sent to the cloud server and saves as the spy of the user by breath after application interface is shown
Sign behavior collection.
3. smart television video method for customizing according to claim 2, which is characterized in that further include after the step D：
The suggestion feedback information that user is directed to shown category video information is received, using multinuclear population clustering algorithm to this
The progress nuclear space mapping of suggestion feedback information and clustering obtain result set, are somebody's turn to do what the result set and cloud server preserved
The characteristic behavior collection of user is compared, and obtains the characteristic information of user, utilizes the characteristic behavior of this feature information update user
Collection.
4. smart television video method for customizing according to claim 2, which is characterized in that further include before the step A：
Middle layer receives the log-on message of user, and the log-on message is sent to cloud server；
Cloud server searches whether there is the user characteristics behavior collection being consistent with the log-on message, and if it exists, will then be searched
To user characteristics behavior collection shown in application interface.
5. smart television video method for customizing according to claim 4, which is characterized in that further include before the step A：
Middle layer obtains device id, and sends it to cloud server；
Cloud server searches whether there is the Family characteristics behavior collection being consistent with the device id, if there are Family characteristics behavior collection
And do not find the user characteristics behavior collection of active user, then by the Family characteristics behavior collection found application interface into
Row display.
6. smart television video method for customizing according to claim 1, which is characterized in that the step C is specially：
C1, middle layer receive the video information returned, set up multinuclear nuclear matrix and carry out core sky to the video information using it
Between map, obtain nuclear space video information；
C2, obtained nuclear space video information is input to as particle in particle cluster algorithm, is commented by iteration and fitness
Valence selects optimal particle；
C3, it is input to selected optimal particle as initial cluster center in k-means clustering algorithms, most using its update
Excellent particle, obtain include multiple category video information cluster result；
C4, the cluster result and the user video customized demand information are compared, obtains customizing with the user video
The most matched category video information of demand information.
7. smart television video method for customizing according to claim 6, which is characterized in that set up multinuclear in the step C1
Nuclear matrix is specially：
Use expression formula forMultinuclear nuclear matrix, whereinFor multiple and different kernel functions,For core letter
Several linear combination coefficient, when givenA different kernel function, then be corresponding withA different nuclear matrix, by optimal
Linear combination coefficient willA nuclear matrix is added, and obtains more preferably multinuclear nuclear matrix。
8. smart television video method for customizing according to claim 6, which is characterized in that the step C2 is specially：
C21, obtained nuclear space video information is input to as particle in particle cluster algorithm, and is calculated by n times each
The fitness of particle obtains N number of primary group；
The fitness of C22, the current fitness by comparing each particle and its desired positions lived through are each to update
The desired positions of particle, the fitness of the desired positions lived through by the current fitness and group that compare each particle come
Update the global desired positions of each particle；
C23, speed and the position that particle is adjusted according to particle cluster algorithm, calculate Colony fitness variance and judge whether it is less than
Predetermined threshold value, if so, selecting optimal particle and executing step C3.
9. smart television video method for customizing according to claim 6, which is characterized in that the step C3 is specially：
C31, using selected optimal particle as the initial cluster center of k-means clustering algorithms, and according to arest neighbors rule
Determine the clustering of corresponding optimal particle；
C32, new cluster centre is calculated according to identified clustering, it is used in combination to replace initial cluster center；
C33, judge cluster centre whether also variation or whether reach maximum iteration, if cluster centre no longer change or
Person reaches maximum iteration, thens follow the steps C34；
C34, judge whether the fitness of new cluster centre is better than previous particle, if so, update optimal particle, is wrapped
Cluster result containing multiple category video information.
10. a kind of smart television video custom-built system, which is characterized in that the system comprises：
Middle layer for receiving user video customized demand information, and will be sent to cloud server after the information preservation；And
Receive the video information that cloud server is returned according to user video customized demand information, utilizes multinuclear population clustering algorithm pair
The video information carries out nuclear space mapping and clustering, obtain include multiple category video information cluster result, by this
Cluster result is compared with the user video customized demand information, is obtained with the user video customized demand information most
The category video information matched；The multinuclear population clustering algorithm has merged multinuclear nuclear matrix algorithm, particle cluster algorithm and k-
Means clustering algorithms；
Cloud server, for according to the corresponding video information of user video customized demand information search received, and should
Video information is back to middle layer；
Application interface display module obtained is regarded for showing with the most matched classification of user video customized demand information
Frequency information.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410115778.7A CN104954873B (en) | 2014-03-26 | 2014-03-26 | A kind of smart television video method for customizing and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410115778.7A CN104954873B (en) | 2014-03-26 | 2014-03-26 | A kind of smart television video method for customizing and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104954873A CN104954873A (en) | 2015-09-30 |
| CN104954873B true CN104954873B (en) | 2018-10-26 |
Family
ID=54169159
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410115778.7A Active CN104954873B (en) | 2014-03-26 | 2014-03-26 | A kind of smart television video method for customizing and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104954873B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106570014B (en) * | 2015-10-09 | 2020-09-25 | 阿里巴巴集团控股有限公司 | Method and apparatus for determining home attribute information of user |
| CN105956113B (en) * | 2016-05-05 | 2019-05-31 | 南京邮电大学 | Video data digging High Dimensional Clustering Analysis method based on particle group optimizing |
| CN108133121A (en) * | 2018-02-24 | 2018-06-08 | 北京科技大学 | The method of piezoelectric transducer port equivalent admittance circuit parameter estimation |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101634996A (en) * | 2009-08-13 | 2010-01-27 | 浙江大学 | Individualized video sequencing method based on comprehensive consideration |
| CN103220555A (en) * | 2013-03-27 | 2013-07-24 | 深圳创维数字技术股份有限公司 | Method, device and system for classifying digital television users |
| EP2677758A1 (en) * | 2012-06-19 | 2013-12-25 | Thomson Licensing | Mind opening content recommending system |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030097196A1 (en) * | 2001-11-13 | 2003-05-22 | Koninklijke Philips Electronics N.V. | Method and apparatus for generating a stereotypical profile for recommending items of interest using item-based clustering |
| US8521773B2 (en) * | 2010-05-25 | 2013-08-27 | Nbcuniversal Media, Llc | System and method for web mining and clustering |
| JP2012204894A (en) * | 2011-03-24 | 2012-10-22 | Toshiba Corp | Information recommendation device |
| CN102780920A (en) * | 2011-07-05 | 2012-11-14 | 上海奂讯通信安装工程有限公司 | Television program recommending method and system |
| CN102999493B (en) * | 2011-09-08 | 2018-07-03 | 北京小度互娱科技有限公司 | A kind of method and apparatus for being used to implement video resource recommendation |
| CN103544206B (en) * | 2013-07-16 | 2017-09-15 | Tcl集团股份有限公司 | A kind of realization method and system of personalized recommendation |
-
2014
- 2014-03-26 CN CN201410115778.7A patent/CN104954873B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101634996A (en) * | 2009-08-13 | 2010-01-27 | 浙江大学 | Individualized video sequencing method based on comprehensive consideration |
| EP2677758A1 (en) * | 2012-06-19 | 2013-12-25 | Thomson Licensing | Mind opening content recommending system |
| CN103220555A (en) * | 2013-03-27 | 2013-07-24 | 深圳创维数字技术股份有限公司 | Method, device and system for classifying digital television users |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104954873A (en) | 2015-09-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20210256403A1 (en) | Recommendation method and apparatus | |
| CN111931062B (en) | Training method and related device of information recommendation model | |
| US10223727B2 (en) | E-commerce recommendation system and method | |
| CN108573041B (en) | Probability matrix decomposition recommendation method based on weighted trust relationship | |
| WO2020147594A1 (en) | Method, system, and device for obtaining expression of relationship between entities, and advertisement retrieval system | |
| CN109360097A (en) | Prediction of Stock Index method, apparatus, equipment and storage medium based on deep learning | |
| WO2019134274A1 (en) | Interest exploration method, storage medium, electronic device and system | |
| CN109902235B (en) | User preference clustering collaborative filtering recommendation algorithm based on bat optimization | |
| WO2023065859A1 (en) | Item recommendation method and apparatus, and storage medium | |
| CN112434213B (en) | Training method of network model, information pushing method and related devices | |
| CN106709076B (en) | Social network recommendation device and method based on collaborative filtering | |
| CN112000763B (en) | Method, device, equipment and medium for determining competition relationship of interest points | |
| CN113761359B (en) | Data packet recommendation method, device, electronic equipment and storage medium | |
| CN104954873B (en) | A kind of smart television video method for customizing and system | |
| Kim et al. | Deep blind image quality assessment by employing FR-IQA | |
| CN112948625A (en) | Film recommendation method based on attribute heterogeneous information network embedding | |
| CN113536105A (en) | Recommendation model training method and device | |
| Yu et al. | Hand gesture recognition based on attentive feature fusion | |
| JP2017201535A (en) | Determination device, learning device, determination method, and determination program | |
| Cao et al. | RetinaMOT: rethinking anchor-free YOLOv5 for online multiple object tracking | |
| CN117435819A (en) | Method, system and storage medium for recommending interest points through space-time dynamic perception | |
| GB2576938A (en) | System and method for improved content discovery | |
| CN114780863B (en) | Project recommendation method and device based on artificial intelligence, computer equipment and medium | |
| Peng et al. | Evaluating deep learning for image classification in adversarial environment | |
| CN115309985A (en) | Fairness evaluation method and AI model selection method of recommendation algorithm |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |