CN106792003A - A kind of intelligent advertisement inserting method, device and server - Google Patents
A kind of intelligent advertisement inserting method, device and server Download PDFInfo
- Publication number
- CN106792003A CN106792003A CN201611224892.9A CN201611224892A CN106792003A CN 106792003 A CN106792003 A CN 106792003A CN 201611224892 A CN201611224892 A CN 201611224892A CN 106792003 A CN106792003 A CN 106792003A
- Authority
- CN
- China
- Prior art keywords
- advertisement
- audio
- matching
- video
- loss function
- 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.)
- Granted
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/20—Servers specifically adapted for the distribution of content, e.g. VOD servers; Operations thereof
- H04N21/25—Management operations performed by the server for facilitating the content distribution or administrating data related to end-users or client devices, e.g. end-user or client device authentication, learning user preferences for recommending movies
- H04N21/266—Channel or content management, e.g. generation and management of keys and entitlement messages in a conditional access system, merging a VOD unicast channel into a multicast channel
- H04N21/2668—Creating a channel for a dedicated end-user group, e.g. insertion of targeted commercials based on end-user profiles
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/70—Information retrieval; Database structures therefor; File system structures therefor of video data
- G06F16/78—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/783—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
- G06F16/7834—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content using audio features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/70—Information retrieval; Database structures therefor; File system structures therefor of video data
- G06F16/78—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/783—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
- G06F16/7844—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content using original textual content or text extracted from visual content or transcript of audio data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/40—Client devices specifically adapted for the reception of or interaction with content, e.g. set-top-box [STB]; Operations thereof
- H04N21/45—Management operations performed by the client for facilitating the reception of or the interaction with the content or administrating data related to the end-user or to the client device itself, e.g. learning user preferences for recommending movies, resolving scheduling conflicts
- H04N21/458—Scheduling content for creating a personalised stream, e.g. by combining a locally stored advertisement with an incoming stream; Updating operations, e.g. for OS modules ; time-related management operations
- H04N21/4586—Content update operation triggered locally, e.g. by comparing the version of software modules in a DVB carousel to the version stored locally
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/80—Generation or processing of content or additional data by content creator independently of the distribution process; Content per se
- H04N21/81—Monomedia components thereof
- H04N21/812—Monomedia components thereof involving advertisement data
Abstract
The invention discloses a kind of intelligent advertisement inserting method, device and server, belong to network communication field.The intelligent advertisement inserting method includes：Object is intercutted in acquisition, and the object that intercuts is for video；The audio-frequency information of object is intercutted in acquisition；The advertisement matched with the audio-frequency information is inquired about in default advertisement base；According to matching result, advertisement is intercutted in the video.By text comparison techniques or nerual network technique, in being embodied as the advertisement of video Auto-matching and intercutting video, the precision for intercutting is high, enhances the relevance of advertisement and video, good advertising effect for intelligent advertisement inserting method of the invention.
Description
Technical field
The present invention relates to network communication field, more particularly to a kind of intelligent advertisement inserting method and device.
Background technology
With the progress developed rapidly with wireless communication technology of internet, Web TV is increasingly subject to the welcome of people,
Current Web TV uses P2P stream media technologys so that user connects faster, and buffer time is shorter, with broadcasting higher
Fluency and viewing experience, therefore, Web TV has attracted large batch of user, correspondingly, the demand of network TV advertisement
Become progressively to increase.The web advertisement of the prior art is typically placed in before Web TV starts, or is interspersed in two streams
Between media, in the form of carousel, will be processed by flow medium digital and be broadcast according to certain with the files in stream media of coding
The program of the autonomous layout of sequence is put forward, playing sequence is usually to refer to the text of certain format or XML play list file
It is fixed.
But, user notes being intended to watch network TV program, and the advertisement for being placed in program beginning is typically taken
Measure is shielded or skipped, or opens webpage in advance, wait advertisement just formal viewing, advertisement putting of the prior art in the past
The advertising results that form is played are very undesirable.
Prior art at least has the following disadvantages：
1st, advertisement has no to associate with the video frequency program of viewing, interesting not strong；
2nd, advertisement is played before being placed in program, maskable property is strong, and advertising results are weak.
The content of the invention
In order to solve problem of the prior art, the invention provides a kind of intelligent advertisement inserting method and device, according to regarding
Frequency content intercuts relevant advertisements, and vivid and interesting improves demonstration effect.The technical scheme is as follows：
On the one hand, the invention provides a kind of intelligent advertisement inserting method, methods described includes：
Object is intercutted in acquisition, and the object that intercuts is for video；
The audio-frequency information of object is intercutted in acquisition；
The advertisement matched with the audio-frequency information is inquired about in default advertisement base；
According to matching result, advertisement is intercutted in the video.
The advertisement matched with the audio-frequency information is inquired about in default advertisement base following two modes：
Mode one is
Be stored with advertisement video in the default advertisement base, and each advertisement video is correspondingly arranged on matching label；
It is described the advertisement matched with the audio-frequency information is inquired about in default advertisement base to include：
Audio-frequency information is converted into text message；
Word segmentation processing is carried out to the text message, participle text is obtained；
Participle text is matched with the label that matches in advertisement base；
According to the matching label that matching is obtained, corresponding advertisement video is obtained.
Mode two is
Audio-frequency information is divided into sub-audio；
Target sub-audio is chosen from the sub-audio, audio collection is constituted；
By audio collection input matching deep neural network model, to search the advertisement matched in advertisement base；
According to the matching result that the model is exported, advertisement corresponding with matching result is obtained.
Intelligent advertisement inserting method in mode two also includes matching deep neural network described in training in advance, including：
Audio collection sample data is obtained, the audio collection sample data is marked with match-type；
Loss function is minimized using stochastic gradient descent method；
The loss function minimized by audio collection sample data and completion, is instructed to the matching deep neural network
Practice, obtain model.
Preferably, the use stochastic gradient descent method minimizes loss function and includes：
According to all weights and loss function of neutral net, the gradient of loss function is obtained using back propagation；
According to the gradient, using stochastic gradient descent method, the weight of neutral net is updated；
The weight of renewal is carried out the iteration of preset times, to minimize loss function.
On the other hand, device is intercutted the invention provides a kind of intelligent advertisement, described device includes：
Object module is intercutted, object is intercutted for obtaining, the object that intercuts is for video；
Audio-frequency module, the audio-frequency information of object is intercutted for obtaining；
Enquiry module, for inquiring about the advertisement matched with the audio-frequency information in default advertisement base；
Module is intercutted, for according to matching result, advertisement being intercutted in the video.
Alternatively, the enquiry module has following two modes to realize：
Mode one
The intelligent advertisement intercuts device also includes advertisement library module, and for default storage advertisement video, each advertisement is regarded
Frequency is correspondingly arranged on matching label；
Accordingly, the enquiry module includes：
Transform subblock, for audio-frequency information to be converted into text message；
Participle submodule, for carrying out word segmentation processing to the text message, obtains participle text；
Matched sub-block, for participle text to be matched with the label that matches in advertisement base；
First acquisition submodule, for the matching label obtained according to matching, obtains corresponding advertisement video.
Mode two
The intelligent advertisement intercuts device also includes pre-training module, and deep neural network is matched for training in advance；
Accordingly, the enquiry module includes：
Segmentation submodule, for audio-frequency information to be divided into sub-audio；
Audio collection submodule, for choosing target sub-audio from the sub-audio, constitutes audio collection；
Network model submodule, for by audio collection input matching deep neural network model, to search advertisement base
The advertisement of middle matching；
Second acquisition submodule, for the matching result exported according to the model, obtains corresponding with matching result wide
Accuse.
The pre-training module includes：
Sample submodule, for obtaining audio collection sample data, the audio collection sample data is marked with match-type；
Submodule is minimized, for minimizing loss function using stochastic gradient descent method；
Study submodule, it is deep to the matching for the loss function minimized by audio collection sample data and completion
Degree neutral net is trained, and obtains model.
Preferably, the minimum submodule includes：
Gradient Unit, for all weights and loss function according to neutral net, is lost using back propagation
The gradient of function；
Weight updating block, for according to the gradient, using stochastic gradient descent method, updating the weight of neutral net；
Iteration unit, the iteration for the weight of renewal to be carried out preset times, to minimize loss function.
On the other hand, the invention provides a kind of intelligent advertisement inter-cut server, the server includes one or more
Intelligent advertisement as described above intercuts device.
What the technical scheme that the present invention is provided was brought has the beneficial effect that：
1) by advertisement as video-frequency playing content is intercutted in the floating window of video page, novel form, viewing property is strong；
2) automatic attaching video-frequency playing content intercuts associated advertisements, vivid and interesting；
3) for compared with conventional ads, demonstration effect is notable.
Brief description of the drawings
Technical scheme in order to illustrate more clearly the embodiments of the present invention, below will be to that will make needed for embodiment description
Accompanying drawing is briefly described, it should be apparent that, drawings in the following description are only some embodiments of the present invention, for
For those of ordinary skill in the art, on the premise of not paying creative work, other can also be obtained according to these accompanying drawings
Accompanying drawing.
Fig. 1 is the flow chart of intelligent advertisement inserting method provided in an embodiment of the present invention；
Fig. 2 is the method flow diagram that advertisement insertion is carried out according to converting text mode provided in an embodiment of the present invention；
Fig. 3 is the method flow diagram that advertisement insertion is carried out according to segmentation sub-audio mode provided in an embodiment of the present invention；
Fig. 4 is the training method flow chart of matching deep neural network provided in an embodiment of the present invention；
Fig. 5 is the method flow diagram for obtaining audio collection sample data to be trained provided in an embodiment of the present invention；
Fig. 6 is that model loss function provided in an embodiment of the present invention minimizes method flow diagram；
Fig. 7 is the module frame chart of advertisement insertion device provided in an embodiment of the present invention；
Fig. 8 is the composition schematic diagram of advertisement insertion server provided in an embodiment of the present invention；
Fig. 9 is the structural representation of neuron in CNN network models provided in an embodiment of the present invention；
Figure 10 is the structure chart of LSTM mnemons in RNN network models provided in an embodiment of the present invention.
Specific embodiment
In order that those skilled in the art more fully understand the present invention program, below in conjunction with the embodiment of the present invention
Accompanying drawing, is clearly and completely described to the technical scheme in the embodiment of the present invention, it is clear that described embodiment is only
The embodiment of a part of the invention, rather than whole embodiments.Based on the embodiment in the present invention, ordinary skill people
The every other embodiment that member is obtained under the premise of creative work is not made, should all belong to the model of present invention protection
Enclose.
It should be noted that term " first ", " in description and claims of this specification and above-mentioned accompanying drawing
Two " it is etc. for distinguishing similar object, without for describing specific order or precedence.It should be appreciated that so using
Data can exchange in the appropriate case, so as to embodiments of the invention described herein can with except illustrating herein or
Order beyond those of description is implemented.Additionally, term " comprising " and " having " and their any deformation, it is intended that cover
Lid is non-exclusive to be included, for example, the process, method, device, product or the equipment that contain series of steps or unit are not necessarily limited to
Those steps or unit clearly listed, but may include not list clearly or for these processes, method, product
Or other intrinsic steps of equipment or unit.
A kind of intelligent advertisement inserting method is provided in one embodiment of the invention, referring to Fig. 1, method flow bag
Include：
Object is intercutted in S101, acquisition.
Specifically, the object that intercuts is Internet video, according to the particular content of Internet video in the present embodiment, in video
During intercut the related advertisement of content therewith, the information of advertisement includes but is not limited to floating window small video advertisement, or word is wide
Accuse etc..
S102, acquisition intercut the audio-frequency information of object.
Specifically, video is the combination of image information and audio-frequency information, and audio-frequency information is obtained from video.
S103, match query advertisement.
Searched in default advertisement base and play the related advertisement of content, the form of the advertisement to current video (audio)
Including but not limited to floating window video, floating window animation and barrage copy.
S104, according to matching result, break for commercialsy.
If matching the advertisement related to currently playing content in advertisement base, broadcasting for related content in video is navigated to
Put at position, intercut the advertisement.
In one embodiment of the invention, there is provided the method that advertisement insertion is carried out according to converting text mode, referring to
Fig. 2, method flow includes：
S201, obtain object video to be broken for commercialsy.
S202, the audio-frequency information for obtaining object video.
S203, audio-frequency information is converted into text message.
Specifically, realized using speech recognition technology：A training set is initially set up, some audios changed are included
(completed by people) with numeral；Then training set is utilized, audio is cut into morpheme piece, found in training set using specific algorithm
In most probable spelling words intellectual.By so training, the parameter for finding builds up voice transformation model.
S204, word segmentation processing, obtain participle text.
Text message to being converted to carries out word segmentation processing, and the foundation of participle is that text message is divided and divided by sentence
Word, obtains a single sentence of sentence, or even obtain independent word to the further participle of single sentence.
The flow of participle, is obtained using elder generation according to audio node segmentation audio after the first voice conversion provided except the present embodiment
To the sub-audio that a sentence is complete, then the mode that each sub-audio is converted into text can equally be realized into the present invention, the present invention is right
Its priority processing sequence is not especially limited.
S205, participle text is matched with the label that matches of advertisement video in advertisement base.
Specifically, an advertisement base is preset with, the advertisement of the cooperative advertising businessman dispensing that is stored with advertisement base, each is wide
Announcement is provided with corresponding matching label, and participle text is matched with the label that matches in advertisement base.If in traversal advertisement base
Matching label, not with current participle text matches, then illustrates currently without the broadcast point for breaking for commercialsy, if matching and working as
The label of preceding participle text matches, then illustrate that current broadcast point has the relevant advertisements that can be intercutted.
S206, according to matching result, obtain corresponding advertisement video.
Specifically, according to matching label and the corresponding relation of advertisement in advertisement base, to match label as index, correspondence is found
Advertisement.
S207, the corresponding audio position of participle text for finding matching, commercial breaks video.
Preserve and the participle text for matching tag match, the corresponding relevant position for finding audio, i.e., with spot announcement
Broadcast point, intercut relevant advertisements in the broadcast point.Such as, the advertisement in non-defective unit shop is provided with advertisement base, matching label is (hard
Fruit or American pistachios or snacks), when audio is played as " you are my American pistachios！" when, the advertisement in non-defective unit shop is matched, current
Floating window is ejected on video page, advertisement video or advertising slogan is shown, the floating window is provided with X button, while being provided with limited time
Closing function.
In one embodiment of the invention, there is provided a kind of side that advertisement insertion is carried out according to segmentation sub-audio mode
Method, referring to Fig. 3, method flow includes：
S31, training matching deep neural network model.
Specific training method flow is as shown in figure 4, methods described flow includes：
S311, obtain audio collection sample data to be trained.The audio collection sample data is made up of audio collection one by one,
Each audio collection is to extract to obtain from a complete audio frequency media, or all audios of whole audio frequency media are divided
Mono- image set of Duan Zucheng.Relatively, target sub-audio composition target audio collection is extracted from audio frequency media to obtain audio collection
Preferred scheme, by extracting representative and importance target sub-audio, effectively eliminate and be unprofitable to match advertisement
Sub-audio, not only alleviates the processing load of neural network model, accelerates the processing speed of neural network model, and removal is dry
Immunity option, improves the accuracy of Audio Matching advertisement, and referring to Fig. 5, the method flow for extracting target sub-audio is as follows：
S3111, obtain audio to be trained；
S3112, by audio segmentation be sub-audio；
S3113, cluster sub-audio, obtain multiple clusters；
The nearest sub-audio of S3114, selected distance cluster centre is used as target sub-audio
S3115, target sub-audio composition audio collection, include sample.
Specifically, the clustering method includes average drifting (Mean Shift) algorithm, Fuzzy C-Means Cluster Algorithm, layer
Secondary clustering algorithm etc..The algorithm principle of average drifting (Mean Shift) algorithm is to randomly choose one in the sample and justify
The heart is o, and radius is the region of h, draws the average value of all sample points in this region, and the sample rate of circle centre position is necessarily than equal
Sample rate at value is small or equal, and average is set into the new center of circle repeats above step, until converging to very dense value
Point；The operation principle of the Fuzzy C-Means Cluster Algorithm is that n sample is divided into c group by algorithm, obtains the cluster of each group
Center, finally allows the object function of non-similarity index to reach minimum, and algorithm assigns being subordinate between 0~1 to each sample point
Degree, each degree classified is belonged to by the value of degree of membership come judgement sample；The present embodiment uses K- means clustering algorithms pair
Sub-audio to be clustered is clustered, and for audio collection X={ x1, x2 ..., xn }, n is sub-audio number, if intending being divided into k
Cluster V={ v1, v2 ..., vk }, first randomly select K object as initial cluster centre, then calculate each object with it is each
Each object, is distributed to the cluster centre nearest apart from it by the distance between individual seed cluster centre.Cluster centre and point
Dispensing their object just represents a cluster, once whole objects are all assigned, the cluster centre of each cluster can basis
Existing object is recalculated in cluster, and this process will be repeated constantly, until cluster centre no longer changes, algorithm terminates,
Cluster obtains the k larger audio class of diversity ratio, correspondingly, one or more sons nearest from cluster centre in each audio class
Audio is that (if cluster centre inherently sub-audio, the sub-audio of cluster centre is target used as target sub-audio
Audio).
S312, the gradient to the loss function of neutral net learn, to minimize loss function.
S313, training network, finally give matching deep neural network model.
Loss function in the S312 is the loss function of deep neural network, loss function and neural network model point
The accuracy close relation of class result, in order to improve the classification accuracy of matching deep neural network model, it is necessary to minimum
Change loss function, specific method is as shown in fig. 6, the method flow of the minimization loss function includes：
The gradient of S3121, back propagation counting loss function：Back propagation (Backpropagation, BP) is one
The algorithm kind being used in combination with optimal method, back propagation to the gradient of all weight calculation loss functions in network,
In vector calculus, the gradient of certain point is pointed in the fastest-rising direction of this scalar field in scalar field, is directional derivative ginseng
Amount；
S3122, gradient is fed back to stochastic gradient descent method：Optimal method is not limited to stochastic gradient descent herein
Method, or gradient descent method, or stochastic parallel gradient descent method；
S3123, renewal weight；
S3124, judge whether to reach the iterations of setting, if reaching, perform S3125, if not up to, by weight
Back propagation is iterated to, i.e., S3121-S3124 is continued executing with the weight for updating；
S3125, completion minimize loss function, and current loss function is the result of minimum.
The iterations being manually specified is drawn by multiple experiment and experience, such as set iteration when test
Number of times is 1000 times, finds that the value for iterating to loss function after 200 times just no longer have dropped in test, then when testing next time
Iterations can be set as 300 times, to save the testing time.
S32, obtain object video to be broken for commercialsy.
S33, the audio-frequency information for obtaining object video.
S34, audio-frequency information is divided into sub-audio.
Specifically, it is segmentation foundation with audio node (the middle pause points of two words in audio), audio-frequency information is divided into
Sub-audio.
S35, selection target sub-audio, constitute audio collection.
Alternatively, all sub-audios composition target audio collection for audio segmentation being obtained, it is preferable that the method by clustering
Extract advantageous in the target sub-audio for carrying out advertisement matching from audio, then target audio collection, institute are constituted by target sub-audio
Clustering method is stated with the sorting procedure in above-mentioned S3113 steps, be will not be repeated here.
S36, audio collection input matching deep neural network model.
S37, according to model export, determine the corresponding advertisement video of target sub-audio.
Specifically, when model is built, advertisement ID lists are preset with, model output parameters pass corresponding with advertisement is set
System, i.e. model output result are advertisement ID, then compare advertisement ID lists, determine corresponding advertisement video.
S38, the position for searching target sub-audio in video, commercial breaks video.
Advertisement video and corresponding sub-audio are preserved, its position in video is determined according to sub-audio, inserted in the position
Broadcast the advertisement.
In one embodiment of the invention, there is provided a kind of intelligent advertisement intercuts device, the module architectures of described device
Referring to Fig. 7, described device is included with lower module：
Object module 710 is intercutted, object is intercutted for obtaining, the object that intercuts is for video；
Audio-frequency module 720, the audio-frequency information of object is intercutted for obtaining；
Enquiry module 730, for inquiring about the advertisement matched with the audio-frequency information in default advertisement base；
Module 740 is intercutted, for according to matching result, advertisement being intercutted in the video.
The advertisement insertion device in one embodiment also includes wherein：Advertisement library module 750, for default storage advertisement
Video, each advertisement video is correspondingly arranged on matching label；
Accordingly, the enquiry module 730 includes：
Transform subblock 731, for audio-frequency information to be converted into text message；
Participle submodule 732, for carrying out word segmentation processing to the text message, obtains participle text；
Matched sub-block 733, for participle text to be matched with the label that matches in advertisement base；
First acquisition submodule 734, for the matching label obtained according to matching, obtains corresponding advertisement video.
Advertisement insertion device in another embodiment also includes：Pre-training module 760, matches deep for training in advance
Degree neutral net；
Correspondingly, the enquiry module 730 includes：
Segmentation submodule 735, for audio-frequency information to be divided into sub-audio；
Audio collection submodule 736, for choosing target sub-audio from the sub-audio, constitutes audio collection；
Network model submodule 737, for by audio collection input matching deep neural network model, to search advertisement
The advertisement matched in storehouse；
Second acquisition submodule 738, for the matching result exported according to the model, obtains corresponding with matching result
Advertisement；
Wherein, the pre-training module 760 includes：
Sample submodule 761, for obtaining audio collection sample data, the audio collection sample data is marked with matching class
Type；
Submodule 762 is minimized, for minimizing loss function using stochastic gradient descent method；
Study submodule 763, for the loss function minimized by audio collection sample data and completion, to the matching
Deep neural network is trained, and obtains model.
Wherein, the minimum submodule 762 includes：
Gradient Unit 7621, for all weights and loss function according to neutral net, is obtained using back propagation
The gradient of loss function；
Weight updating block 7622, for according to the gradient, using stochastic gradient descent method, updating the power of neutral net
Weight；
Iteration unit 7623, the iteration for the weight of renewal to be carried out preset times, to minimize loss function.
In one embodiment of the invention, there is provided a kind of intelligent advertisement inter-cut server, referring to Fig. 8, the service
Device intercuts device including one or more above-mentioned intelligent advertisements.
It should be noted that：Above-described embodiment provide Internet of Things control device carry out unified management control when, only with
The division of above-mentioned each functional module is carried out for example, in practical application, as needed can distribute by not above-mentioned functions
With functional module complete, will the internal structure of advertisement insertion device be divided into different functional modules, retouched with completing the above
The all or part of function of stating.In addition, the advertisement insertion device embodiment of the present embodiment offer and above-described embodiment offer
Advertisement cut-in method belongs to same design, and it implements process and refers to embodiment of the method, repeats no more here.
Embodiment of the method provided in an embodiment of the present invention can be filled in mobile terminal, terminal or similar computing
Middle execution is put, in one embodiment of the invention, as a example by running on computer terminals, the terminal can include RF
(Radio Frequency, radio frequency) circuit, include the memory of one or more computer-readable recording mediums, defeated
Enter unit, display unit, sensor, voicefrequency circuit, WiFi (wireless fidelity, Wireless Fidelity) module, include one
The part such as individual or more than one processing core processor and power supply.Wherein：
RF circuits can be used to receiving and sending messages or communication process in, the reception and transmission of signal, especially, by the descending of base station
After information is received, one or the treatment of more than one processor are transferred to；In addition, up data is activation will be related to base station.It is logical
Often, RF circuits include but is not limited to antenna, at least one amplifier, tuner, one or more oscillators, subscriber identity module
(SIM) card, transceiver, coupler, LNA (Low Noise Amplifier, low-noise amplifier), duplexer etc..Additionally,
RF circuits can also be communicated by radio communication with network and other equipment.The radio communication can use any communication standard
Or agreement, including but not limited to GSM (Global System ofMobile communication, global system for mobile telecommunications system
System), GPRS (General Packet Radio Service, general packet radio service), CDMA (Code Division
Multiple Access, CDMA), WCDMA (Wideband Code Division Multiple Access, broadband code
Point multiple access), LTE (Long Term Evolution, Long Term Evolution), Email, SMS (Short Messaging
Service, Short Message Service) etc..
Memory can be used to store software program and module, and processor is by running software program of the storage in memory
And module, so as to perform various function application and data processing.Memory can mainly include storing program area and storage number
According to area, wherein, the application program that storing program area can be needed for storage program area, function (broadcast by such as sound-playing function, image
Playing function etc.) etc.；Storage data field can be stored and use created data (such as voice data, phone directory etc.) according to terminal
Deng.Additionally, memory can include high-speed random access memory, nonvolatile memory, for example, at least one can also be included
Individual disk memory, flush memory device or other volatile solid-state parts.Correspondingly, memory can also include storage
Device controller, to provide the access of processor and input block to memory.
Input block can be used to receive the numeral or character information of input, and generation is set and function control with user
Relevant keyboard, mouse, action bars, optics or trace ball signal input.Specifically, input block may include Touch sensitive surface with
And other input equipments.Touch sensitive surface, also referred to as touch display screen or Trackpad, can collect user thereon or neighbouring touch
(such as user uses any suitable objects such as finger, stylus or annex attached on Touch sensitive surface or in Touch sensitive surface to touch operation
Near operation), and corresponding attachment means are driven according to formula set in advance.Optionally, Touch sensitive surface may include to touch inspection
Survey two parts of device and touch controller.Wherein, touch detecting apparatus detect the touch orientation of user, and detect touch operation
The signal for bringing, transmits a signal to touch controller；Touch controller receives touch information from touch detecting apparatus, and will
It is converted into contact coordinate, then gives processor, and the order sent of receiving processor and can be performed.Furthermore, it is possible to adopt
Touch sensitive surface is realized with polytypes such as resistance-type, condenser type, infrared ray and surface acoustic waves.Except Touch sensitive surface, input is single
Unit can also include other input equipments.Specifically, other input equipments can include but is not limited to physical keyboard, function key
One or more in (such as volume control button, switch key etc.), trace ball, mouse, action bars etc..
Display unit can be used for display by the information of user input or be supplied to the information of user and the various figures of terminal
Shape user interface, these graphical user interface can be made up of figure, text, icon, video and its any combination.Display is single
Unit may include display panel, optionally, can use LCD (Liquid Crystal Display, liquid crystal display), OLED
Forms such as (Organic Light-Emitting Diode, Organic Light Emitting Diode) configures display panel.Further, touch
Sensitive surfaces can cover display panel, when Touch sensitive surface is detected thereon or after neighbouring touch operation, send to processor with
Determine the type of touch event, it is defeated according to the type of touch event to provide corresponding vision on a display panel with preprocessor
Go out.Although in the present embodiment, Touch sensitive surface is that input and input function are realized as two independent parts with display panel,
But in some embodiments it is possible to by Touch sensitive surface and display panel it is integrated and realize input and output function.
Terminal may also include at least one sensor, such as optical sensor, motion sensor and other sensors.Specifically
Ground, optical sensor may include ambient light sensor and proximity transducer, wherein, ambient light sensor can be according to the bright of ambient light
The brightness of display panel secretly is adjusted, proximity transducer can close display panel and/or backlight when terminal is moved in one's ear.
Used as one kind of motion sensor, (generally three axles) acceleration is big in the detectable all directions of Gravity accelerometer
It is small, size and the direction of gravity are can detect that when static, can be used for application (such as horizontal/vertical screen switching, the phase of identification terminal attitude
Close game, magnetometer pose calibrating), Vibration identification correlation function (such as pedometer, tap) etc.；Be can also configure as terminal
The other sensors such as gyroscope, barometer, hygrometer, thermometer, infrared ray sensor, will not be repeated here.
Voicefrequency circuit, loudspeaker, microphone can provide the COBBAIF between user and terminal.Voicefrequency circuit will can be received
Electric signal after the voice data conversion arrived, is transferred to loudspeaker, and being converted to voice signal by loudspeaker exports；On the other hand,
The voice signal of collection is converted to electric signal by microphone, and voice data is converted to after being received by voicefrequency circuit, then by audio number
After being processed according to output processor, through RF circuits being sent to such as another terminal, or by voice data export to memory with
Just further treatment.Voicefrequency circuit is also possible that earphone jack, to provide the communication of peripheral hardware earphone and terminal.
WiFi belongs to short range wireless transmission technology, terminal can help user to send and receive e-mail by WiFi module,
Browse webpage and access streaming video etc., it has provided the user wireless broadband internet and has accessed.It is understood that WiFi
Module is simultaneously not belonging to must be configured into for terminal, can be omitted in the essential scope for do not change invention as needed completely.
Processor is the control centre of terminal, using various interfaces and the various pieces of the whole terminal of connection, is passed through
Operation performs software program and/or module of the storage in memory, and calls data of the storage in memory, performs
The various functions and processing data of terminal, so as to carry out integral monitoring to terminal.Optionally, processor may include one or more
Processing core；Preferably, processor can integrated application processor and modem processor, wherein, application processor is mainly located
Reason operating system, user interface and application program etc., modem processor mainly processes radio communication.It is understood that
Above-mentioned modem processor can not also be integrated into processor.
Terminal also includes the power supply (such as battery) powered to all parts, it is preferred that power supply can be by power management
System is logically contiguous with processor, so as to realize the work(such as management charging, electric discharge and power managed by power-supply management system
Energy.Power supply can also include one or more direct current or AC power, recharging system, power failure detection circuit,
The random component such as power supply changeover device or inverter, power supply status indicator.
Although not shown, terminal can also will not be repeated here including camera, bluetooth module etc..Specifically in this implementation
In example, the display unit of terminal is touch-screen display, and terminal also includes memory, and one or more than one journey
Sequence, one of them or more than one program storage is configured to by one or more than one processor in memory
Execution states one or more than one program bag contains the instruction for being used for carrying out following operation：
Object is intercutted in acquisition, and the object that intercuts is for video；
The audio-frequency information of object is intercutted in acquisition；
The advertisement matched with the audio-frequency information is inquired about in default advertisement base；
According to matching result, advertisement is intercutted in the video.
Further, be stored with advertisement video in the default advertisement base, and each advertisement video is correspondingly arranged on matching
Label；
Specifically, the processor of terminal is additionally operable to perform the instruction of following operation：
Audio-frequency information is converted into text message；
Word segmentation processing is carried out to the text message, participle text is obtained；
Participle text is matched with the label that matches in advertisement base；
According to the matching label that matching is obtained, corresponding advertisement video is obtained.
Another way is that the processor of terminal is additionally operable to perform the instruction of following operation：Audio-frequency information is divided into son
Audio；
Target sub-audio is chosen from the sub-audio, audio collection is constituted；
By audio collection input matching deep neural network model, to search the advertisement matched in advertisement base；
According to the matching result that the model is exported, advertisement corresponding with matching result is obtained.
Specifically, the processor of terminal is additionally operable to perform the instruction of following operation：Depth nerve is matched described in training in advance
Network, including：
Audio collection sample data is obtained, the audio collection sample data is marked with match-type；
Loss function is minimized using stochastic gradient descent method；
The loss function minimized by audio collection sample data and completion, is instructed to the matching deep neural network
Practice, obtain model.
Specifically, the processor of terminal is additionally operable to perform the instruction of following operation：
According to all weights and loss function of neutral net, the gradient of loss function is obtained using back propagation；
According to the gradient, using stochastic gradient descent method, the weight of neutral net is updated；
The weight of renewal is carried out the iteration of preset times, to minimize loss function.
By the description of embodiment of above, those skilled in the art can be understood that regarding for present invention offer
Frequency marking topic generation technique scheme can add the mode of required general hardware platform to realize by software, naturally it is also possible to by hard
Part, but the former is more preferably implementation method in many cases.Based on such understanding, technical scheme substantially or
Say that the part contributed to prior art can be embodied in the form of software product, computer software product storage exists
In one storage medium (such as ROM/RAM, magnetic disc, CD), including some instructions are used to so that a station terminal equipment (can be hand
Machine, computer, server, or network equipment etc.) perform method described in each embodiment of the invention.
In one embodiment of the invention, there is provided a kind of computer-readable recording medium, computer-readable storage
Medium can be the computer-readable recording medium included in the memory in above-described embodiment；Can also be individualism,
Without the computer-readable recording medium allocated into terminal.Computer-readable recording medium storage has one or more than one journey
Sequence, the method that or more than one program are used for performing advertisement insertion by or more than one processor is described
Method includes：
Object is intercutted in acquisition, and the object that intercuts is for video；
The audio-frequency information of object is intercutted in acquisition；
The advertisement matched with the audio-frequency information is inquired about in default advertisement base；
According to matching result, advertisement is intercutted in the video.
The advertisement matched with the audio-frequency information is inquired about in default advertisement base following two modes：
Mode one is
Be stored with advertisement video in the default advertisement base, and each advertisement video is correspondingly arranged on matching label；
It is described the advertisement matched with the audio-frequency information is inquired about in default advertisement base to include：
Audio-frequency information is converted into text message；
Word segmentation processing is carried out to the text message, participle text is obtained；
Participle text is matched with the label that matches in advertisement base；
According to the matching label that matching is obtained, corresponding advertisement video is obtained.
Mode two is
Audio-frequency information is divided into sub-audio；
Target sub-audio is chosen from the sub-audio, audio collection is constituted；
By audio collection input matching deep neural network model, to search the advertisement matched in advertisement base；
According to the matching result that the model is exported, advertisement corresponding with matching result is obtained.
Intelligent advertisement inserting method in mode two also includes matching deep neural network described in training in advance, including：
Audio collection sample data is obtained, the audio collection sample data is marked with match-type；
Loss function is minimized using stochastic gradient descent method；
The loss function minimized by audio collection sample data and completion, is instructed to the matching deep neural network
Practice, obtain model.
Preferably, the use stochastic gradient descent method minimizes loss function and includes：
According to all weights and loss function of neutral net, the gradient of loss function is obtained using back propagation；
According to the gradient, using stochastic gradient descent method, the weight of neutral net is updated；
The weight of renewal is carried out the iteration of preset times, to minimize loss function.
In one embodiment of the invention, matching depth nerve is obtained using CNN (convolutional neural networks) model framework
Network model, the input data handling process of CNN models includes：
Firstth, the extraction conditions of extraction target sub-audio in audio are defined；
Secondth, the video (audio) for treating commercial breaks extracts the target sub-audio for meeting said extracted condition；
3rd, for the audio collection that each is made up of target sub-audio, its sub-audio is pressed into the arrangement of category attribution degree descending,
The category attribution degree is defined as：
Category attribution degree=(degree of degree/node in node circle in artwork) * (degree/circle in node circle
Figure maximal degree).
Sample data is spliced into three-dimensional array, three dimensions are respectively circle, image member and data and lead to from outside to inside
Road, the number of members of each circle must be equal in this three-dimensional array, and this quantity is set into M, circle of the number of members more than K
The M most preceding member data of member of interception ranking, the circle of lazy weight M is supplied with 0.
The architecture design of the CNN models is as follows：Comprising two 2D convolutional layers (convolution2d_1,
Convolution2d_2), two full articulamentums (dense_1, dense_2), using convolution2d_input to nerve
Network is input into, and convolution2d_input_1 (InputLayer) is the input layer of neutral net, the nothing in this layer
Any computing, only defines the size and type of input data, therefore, output output quantities do not change.
Convolution2D is the convolutional layer of 2 dimensions, and convolutional layer comes simple few model parameter and data fortune by parameter sharing
Calculate, the major parameter of convolutional layer includes：A. convolution nuclear volume, each convolution kernel one feature map of correspondence, the number of convolution kernel
Amount can show that the quantity of feature map is 64 in the present embodiment by the quantity of feature map；B. convolution kernel it is long,
Width, the convolution kernel is a rectangle, it is necessary to specified length and width, the volume of convolution kernel is 3x3 in the present embodiment；C. step-length, refers to
Step-length of the convolution kernel in translation, because convolution kernel is 2 dimension datas, correspondingly, step-length is the array that length is 2,
Such as (1,1), the neuron of convolutional layer is shared using weight (weights), and each neuron weights quantity=convolution kernel is long
X convolution kernels are wide.
Activation is the activation primitive of neuron, and in neutral net, in addition to last layer of output, remaining is appointed
What neuron has activation primitive, and the activation primitive of each all of neuron of layer is identical, and the neuron of different layers has
Different activation primitives.There is a weight on each input side of neuron, and each neuron has a biasing (bias),
In the present embodiment, using activation primitive ReLu, function is defined as g (z)=max { 0, z }.
MaxPooling2D is an operation for 2-D data, specially takes the greatest measure output in a rectangle,
The major parameter of the MaxPooling2D includes：A.Pool sizes, refer to a rectangle, such as 3x3；B. step-length, refers to every time
Mobile length, such as (3,3).
The purpose of Dropout is that over-fitting is one of most common problem of machine learning, for retouching in order to prevent over-fitting
Performance of the model on training set is stated far better than the performance on test set.If that is, a model over-fitting, that
It performs well in training, but effect is very different when doing actual prediction with new data, the major parameter of the Dropout
Including：Parameter p：Value between one 0 to 1, represents probability, when training pattern, at random by the input of this layer (namely
Above one layer of output) 0, such as p=0.2 are set as according to p probability, that is just random to be set as 0 by 20% input node data, but
In forecast period, the layer does not do any operation.
Flatten is acted on and is flattened into two-dimensional array one-dimensional, such as be changed into [1,2,3,4] [[1,2], [3,4]].
Dense be full articulamentum, in general, Hidden layer are exactly full articulamentum, if neuron as shown in figure 9,
Operational formula is as follows：
Output=g (z), wherein, the g (z) is activation primitive, is specifically defined as above, be will not be repeated here；
Z=∑sjwjxj+ b, wherein, xiIt is i-th input, wiIt is i-th weight of input, b is offset threshold.
Because being many classification problems, each video circle belongs to a classification, and output layer is softmax, loss function
(Loss function) uses stochastic gradient descent method from classification cross entropy (categorical cross entropy)
(SGD) learning model parameter, learning process is as above trained described in the step of matching neutral net.
In another embodiment of the present invention, matching depth god is obtained using RNN (Recognition with Recurrent Neural Network) model framework
Through network model, it is descending arrangement to be carried out to the member of each circle also according to category attribution degree, with CNN with CNN identicals
Unlike, in RNN, arrangement obtains a sequence on member data, in sequence each correspond to user
Personal data, the corresponding sequence of each circle allows different length, that is to say, that the quantity of circle member can be with inconsistent.
The architecture design of the RNN models is as follows：Comprising three LSTM layers (lstm_1, lstm_2, lstm_3) and two
Full articulamentum (dense_1, dense_2).
RNN neutral nets are input into using lstm_input to neutral net, lstm_input_1 (InputLayer)
It is the input layer of RNN neutral nets, without any computing in this layer, only defines the size and type of input data, because
This, output output quantities are not changed, and the structure of LSTM mnemons is shown in Figure 10.
Full articulamentum in RNN neutral nets and prevent over-fitting layer respectively with the full articulamentum of CNN neutral nets and prevent
Only the definition of over-fitting layer is identical, will not be repeated here.
The embodiments of the present invention are for illustration only, and the quality of embodiment is not represented.
One of ordinary skill in the art will appreciate that realizing that all or part of step of above-described embodiment can be by hardware
To complete, it is also possible to instruct the hardware of correlation to complete by program, described program can be stored in a kind of computer-readable
In storage medium, storage medium mentioned above can be read-only storage, disk or CD etc..
The foregoing is only presently preferred embodiments of the present invention, be not intended to limit the invention, it is all it is of the invention spirit and
Within principle, any modification, equivalent substitution and improvements made etc. should be included within the scope of the present invention.
Claims (10)
1. a kind of intelligent advertisement inserting method, it is characterised in that methods described includes：
Object is intercutted in acquisition, and the object that intercuts is for video；
The audio-frequency information of object is intercutted in acquisition；
The advertisement matched with the audio-frequency information is inquired about in default advertisement base；
According to matching result, advertisement is intercutted in the video.
2. method according to claim 1, it is characterised in that
Be stored with advertisement video in the default advertisement base, and each advertisement video is correspondingly arranged on matching label；
It is described the advertisement matched with the audio-frequency information is inquired about in default advertisement base to include：
Audio-frequency information is converted into text message；
Word segmentation processing is carried out to the text message, participle text is obtained；
Participle text is matched with the label that matches in advertisement base；
According to the matching label that matching is obtained, corresponding advertisement video is obtained.
3. method according to claim 1, it is characterised in that described to be inquired about in default advertisement base and audio letter
The advertisement for ceasing matching includes：
Audio-frequency information is divided into sub-audio；
Target sub-audio is chosen from the sub-audio, audio collection is constituted；
By audio collection input matching deep neural network model, to search the advertisement matched in advertisement base；
According to the matching result that the model is exported, advertisement corresponding with matching result is obtained.
4. method according to claim 3, it is characterised in that also including matching deep neural network described in training in advance,
Including：
Audio collection sample data is obtained, the audio collection sample data is marked with match-type；
Loss function is minimized using stochastic gradient descent method；
The loss function minimized by audio collection sample data and completion, is trained to the matching deep neural network,
Obtain model.
5. method according to claim 4, it is characterised in that the use stochastic gradient descent method minimizes loss function
Including：
According to all weights and loss function of neutral net, the gradient of loss function is obtained using back propagation；
According to the gradient, using stochastic gradient descent method, the weight of neutral net is updated；
The weight of renewal is carried out the iteration of preset times, to minimize loss function.
6. a kind of intelligent advertisement intercuts device, it is characterised in that described device includes：
Object module is intercutted, object is intercutted for obtaining, the object that intercuts is for video；
Audio-frequency module, the audio-frequency information of object is intercutted for obtaining；
Enquiry module, for inquiring about the advertisement matched with the audio-frequency information in default advertisement base；
Module is intercutted, for according to matching result, advertisement being intercutted in the video.
7. device according to claim 6, it is characterised in that also include：
Advertisement library module, for default storage advertisement video, each advertisement video is correspondingly arranged on matching label；
The enquiry module includes：
Transform subblock, for audio-frequency information to be converted into text message；
Participle submodule, for carrying out word segmentation processing to the text message, obtains participle text；
Matched sub-block, for participle text to be matched with the label that matches in advertisement base；
First acquisition submodule, for the matching label obtained according to matching, obtains corresponding advertisement video.
8. device according to claim 6, it is characterised in that also include：
Pre-training module, deep neural network is matched for training in advance；
The enquiry module includes：
Segmentation submodule, for audio-frequency information to be divided into sub-audio；
Audio collection submodule, for choosing target sub-audio from the sub-audio, constitutes audio collection；
Network model submodule, for by audio collection input matching deep neural network model, to search in advertisement base
The advertisement matched somebody with somebody；
Second acquisition submodule, for the matching result exported according to the model, obtains advertisement corresponding with matching result；
The pre-training module includes：
Sample submodule, for obtaining audio collection sample data, the audio collection sample data is marked with match-type；
Submodule is minimized, for minimizing loss function using stochastic gradient descent method；
Study submodule, for the loss function minimized by audio collection sample data and completion, to the matching depth god
It is trained through network, obtains model.
9. device according to claim 8, it is characterised in that the minimum submodule includes：
Gradient Unit, for all weights and loss function according to neutral net, loss function is obtained using back propagation
Gradient；
Weight updating block, for according to the gradient, using stochastic gradient descent method, updating the weight of neutral net；
Iteration unit, the iteration for the weight of renewal to be carried out preset times, to minimize loss function.
10. a kind of intelligent advertisement inter-cut server, it is characterised in that including one or more as any one in claim 6-9
Intelligent advertisement described in intercuts device.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611224892.9A CN106792003B (en) | 2016-12-27 | 2016-12-27 | Intelligent advertisement insertion method and device and server |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611224892.9A CN106792003B (en) | 2016-12-27 | 2016-12-27 | Intelligent advertisement insertion method and device and server |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106792003A true CN106792003A (en) | 2017-05-31 |
| CN106792003B CN106792003B (en) | 2020-04-14 |
Family
ID=58920982
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611224892.9A Active CN106792003B (en) | 2016-12-27 | 2016-12-27 | Intelligent advertisement insertion method and device and server |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106792003B (en) |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108184153A (en) * | 2017-12-29 | 2018-06-19 | 伟乐视讯科技股份有限公司 | A kind of advertisement insertion system to match with video content and method |
| CN108364558A (en) * | 2018-03-09 | 2018-08-03 | 南京信息职业技术学院 | A kind of agricultural sand table system |
| CN109978145A (en) * | 2019-03-29 | 2019-07-05 | 联想(北京)有限公司 | A kind of processing method and processing device |
| CN110290404A (en) * | 2019-07-31 | 2019-09-27 | 安徽抖范视频科技有限公司 | A kind of inserting method of video ads and intercut device |
| CN110401855A (en) * | 2018-04-25 | 2019-11-01 | 腾讯科技（深圳）有限公司 | Information displaying method, processing platform, calculates equipment and storage medium at device |
| CN111050194A (en) * | 2019-12-02 | 2020-04-21 | 北京奇艺世纪科技有限公司 | Video sequence processing method, video sequence processing device, electronic equipment and computer readable storage medium |
| CN111104370A (en) * | 2019-12-18 | 2020-05-05 | 北京大龙得天力广告传媒有限公司 | Advertisement video storage system and method |
| CN111754267A (en) * | 2020-06-29 | 2020-10-09 | 蚌埠科睿达机械设计有限公司 | Data processing method and system based on block chain |
| WO2021185184A1 (en) * | 2020-03-20 | 2021-09-23 | 华为技术有限公司 | Content recommendation method and apparatus, electronic device, and storage medium |
| CN113434794A (en) * | 2021-06-23 | 2021-09-24 | 平安国际智慧城市科技股份有限公司 | Element carousel method, device, equipment and medium |
| WO2022093453A1 (en) * | 2020-10-30 | 2022-05-05 | Google Llc | Transforming data from streaming media |
| CN115065837A (en) * | 2022-05-13 | 2022-09-16 | 咪咕视讯科技有限公司 | Video insertion method, device and equipment and computer readable storage medium |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101072340A (en) * | 2007-06-25 | 2007-11-14 | 孟智平 | Method and system for adding advertising information in flow media |
| WO2012167568A1 (en) * | 2011-11-23 | 2012-12-13 | 华为技术有限公司 | Video advertisement broadcasting method, device and system |
| CN103366740A (en) * | 2012-03-27 | 2013-10-23 | 联想(北京)有限公司 | Voice command recognition method and voice command recognition device |
| CN104853223A (en) * | 2015-04-29 | 2015-08-19 | 小米科技有限责任公司 | Video stream intercutting method and terminal equipment |
| CN104951965A (en) * | 2015-06-26 | 2015-09-30 | 深圳市腾讯计算机系统有限公司 | Advertisement delivery method and device |
| CN104992347A (en) * | 2015-06-17 | 2015-10-21 | 北京奇艺世纪科技有限公司 | Video matching advertisement method and device |
| CN106162328A (en) * | 2015-04-28 | 2016-11-23 | 天脉聚源（北京）科技有限公司 | A kind of video synchronizing information methods of exhibiting and system |
-
2016
- 2016-12-27 CN CN201611224892.9A patent/CN106792003B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101072340A (en) * | 2007-06-25 | 2007-11-14 | 孟智平 | Method and system for adding advertising information in flow media |
| WO2012167568A1 (en) * | 2011-11-23 | 2012-12-13 | 华为技术有限公司 | Video advertisement broadcasting method, device and system |
| CN103366740A (en) * | 2012-03-27 | 2013-10-23 | 联想(北京)有限公司 | Voice command recognition method and voice command recognition device |
| CN106162328A (en) * | 2015-04-28 | 2016-11-23 | 天脉聚源（北京）科技有限公司 | A kind of video synchronizing information methods of exhibiting and system |
| CN104853223A (en) * | 2015-04-29 | 2015-08-19 | 小米科技有限责任公司 | Video stream intercutting method and terminal equipment |
| CN104992347A (en) * | 2015-06-17 | 2015-10-21 | 北京奇艺世纪科技有限公司 | Video matching advertisement method and device |
| CN104951965A (en) * | 2015-06-26 | 2015-09-30 | 深圳市腾讯计算机系统有限公司 | Advertisement delivery method and device |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108184153A (en) * | 2017-12-29 | 2018-06-19 | 伟乐视讯科技股份有限公司 | A kind of advertisement insertion system to match with video content and method |
| CN108364558A (en) * | 2018-03-09 | 2018-08-03 | 南京信息职业技术学院 | A kind of agricultural sand table system |
| CN110401855A (en) * | 2018-04-25 | 2019-11-01 | 腾讯科技（深圳）有限公司 | Information displaying method, processing platform, calculates equipment and storage medium at device |
| CN109978145A (en) * | 2019-03-29 | 2019-07-05 | 联想(北京)有限公司 | A kind of processing method and processing device |
| CN110290404A (en) * | 2019-07-31 | 2019-09-27 | 安徽抖范视频科技有限公司 | A kind of inserting method of video ads and intercut device |
| CN111050194A (en) * | 2019-12-02 | 2020-04-21 | 北京奇艺世纪科技有限公司 | Video sequence processing method, video sequence processing device, electronic equipment and computer readable storage medium |
| CN111104370A (en) * | 2019-12-18 | 2020-05-05 | 北京大龙得天力广告传媒有限公司 | Advertisement video storage system and method |
| WO2021185184A1 (en) * | 2020-03-20 | 2021-09-23 | 华为技术有限公司 | Content recommendation method and apparatus, electronic device, and storage medium |
| CN111754267A (en) * | 2020-06-29 | 2020-10-09 | 蚌埠科睿达机械设计有限公司 | Data processing method and system based on block chain |
| WO2022093453A1 (en) * | 2020-10-30 | 2022-05-05 | Google Llc | Transforming data from streaming media |
| CN113434794A (en) * | 2021-06-23 | 2021-09-24 | 平安国际智慧城市科技股份有限公司 | Element carousel method, device, equipment and medium |
| CN115065837A (en) * | 2022-05-13 | 2022-09-16 | 咪咕视讯科技有限公司 | Video insertion method, device and equipment and computer readable storage medium |
| CN115065837B (en) * | 2022-05-13 | 2023-10-17 | 咪咕视讯科技有限公司 | Video inter-cut method, device, equipment and computer readable storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106792003B (en) | 2020-04-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106792003A (en) | A kind of intelligent advertisement inserting method, device and server | |
| CN106779073B (en) | Media information classification method and device based on deep neural network | |
| CN106845390B (en) | Video title generation method and device | |
| CN108875781B (en) | Label classification method and device, electronic equipment and storage medium | |
| CN107943860A (en) | The recognition methods and device that the training method of model, text are intended to | |
| CN110598046B (en) | Artificial intelligence-based identification method and related device for title party | |
| CN108280458A (en) | Group relation kind identification method and device | |
| CN109784424A (en) | A kind of method of image classification model training, the method and device of image procossing | |
| CN110162770A (en) | A kind of word extended method, device, equipment and medium | |
| CN109241431A (en) | A kind of resource recommendation method and device | |
| CN110232696A (en) | A kind of method of image region segmentation, the method and device of model training | |
| CN114416927A (en) | Intelligent question and answer method, device, equipment and storage medium | |
| CN112101329B (en) | Video-based text recognition method, model training method and model training device | |
| CN104217717A (en) | Language model constructing method and device | |
| CN108434740A (en) | A kind of method and device that policy information determines | |
| CN110166828A (en) | A kind of method for processing video frequency and device | |
| CN109561211B (en) | Information display method and mobile terminal | |
| CN111882362A (en) | Artificial intelligence advertisement delivery system based on 5G communication network | |
| CN108021572A (en) | Return information recommends method and apparatus | |
| CN108334539A (en) | Object recommendation method, mobile terminal and computer readable storage medium | |
| CN111368171B (en) | Keyword recommendation method, related device and storage medium | |
| CN113378556A (en) | Method and device for extracting text keywords | |
| CN110209810A (en) | Similar Text recognition methods and device | |
| CN108289057A (en) | Data sharing method, system and mobile terminal | |
| CN106874936A (en) | Image propagates monitoring method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |