CN106021496A

CN106021496A - Video search method and video search device

Info

Publication number: CN106021496A
Application number: CN201610341232.2A
Authority: CN
Inventors: 马宏; 王峰; 匡涛; 任晓楠
Original assignee: Hisense Group Co Ltd
Current assignee: Hisense Group Co Ltd
Priority date: 2016-05-19
Filing date: 2016-05-19
Publication date: 2016-10-12

Abstract

The invention provides a video search method and a video search device. The video search method comprises the steps of receiving video search information, wherein the video search information comprises role information of a target role, and the role information is used for identifying different roles; acquiring a video clip corresponding to the target role according to the role information of the target role; and generating a search video corresponding to the target role according to the video clip corresponding to the target role. The video search method and the video search device provided by the invention can enable users to watch their concern role video clips in a video, so as to meet the individual requirements of the users for video search.

Description

Video searching method and video searching apparatus

Technical field

The present invention relates to field of video processing, particularly relate to a kind of video searching method and video searching apparatus.

Background technology

Along with development and the lifting of the network bandwidth of Internet technology, people carry out video playback and viewing by the network media more and more.

But, because the data volume of Internet video becomes geometric growth, in massive video the most on the internet, fast searching becomes a stubborn problem to the video meeting user's request.At present, the main path of video search includes that search engine is searched for and video display client-side search etc. indirectly.Searching method is mainly based upon the keyword searches such as film name, performer and the director of video, or video search based on video display classification, and the unit represented also is that video is overall, and as TV play diversity is play, film is according to whole footage etc..

But, quickening along with modern life rhythm, the time that user can spend in video display viewing is fewer and feweri, it is overall that more user is no longer want to watch video, and be intended to quickly watch the role's fragment oneself paid close attention in video, therefore, existing video searching method can not meet the individual demand of user.

Summary of the invention

The present invention provides a kind of video searching method and video searching apparatus, it is intended to provides the user the role's fragment oneself paid close attention in target video, meets user's individual demand to video search.

First aspect, the present invention provides a kind of video searching method, including:

Receiving video searching information, wherein, video searching information includes the Role Information of target roles, and Role Information is used for identifying different role；

Role Information according to target roles obtains the video segment that target roles is corresponding；

The search video corresponding to target roles is generated according to the video segment that target roles is corresponding.

Second aspect, the present invention provides a kind of video searching apparatus, including:

Receiver module, is used for receiving video searching information, and wherein, video searching information includes the Role Information of target roles, and Role Information is used for identifying different role；

Video acquiring module, obtains, for the Role Information according to target roles, the video segment that target roles is corresponding；

Video-splicing module, generates the search video corresponding to target roles for the video segment corresponding according to target roles.

The third aspect, the present invention provides a kind of video searching apparatus, including:

Receptor, is used for receiving video searching information, and wherein, video searching information includes the Role Information of target roles, and Role Information is used for identifying different role；

Memorizer, is used for storing program；

Processor, for performing the program of memorizer storage, obtains, with the Role Information according to target roles, the video segment that described target roles is corresponding, and generates, according to the video segment that described target roles is corresponding, the search video that described target roles is corresponding.

The video searching method that the present invention provides, first receives video searching information, and wherein, video searching information includes the Role Information of target roles, and Role Information is used for identifying different role；Role Information further according to target roles obtains all video segments that target roles is corresponding；The all video segments corresponding finally according to target roles obtain the video corresponding to target roles.User so can be allowed to watch the video segment corresponding to the role oneself paid close attention in target video, eliminate user oneself manual search, the process of adjustment playing progress rate, the viewing being effectively increased user is experienced, and meets user's individual demand to video search.

Accompanying drawing explanation

In order to be illustrated more clearly that the embodiment of the present invention or technical scheme of the prior art, the accompanying drawing used required in embodiment or description of the prior art will be briefly described below, apparently, accompanying drawing in describing below is some embodiments of the present invention, for those of ordinary skill in the art, on the premise of not paying creative work, it is also possible to obtain other accompanying drawing according to these accompanying drawings.

Figure 1A is the schematic flow sheet of a kind of video searching method that the embodiment of the present invention one provides；

Figure 1B is the schematic flow sheet of the another kind of video searching method that the embodiment of the present invention one provides；

Fig. 1 C is the schematic flow sheet of the third video searching method that the embodiment of the present invention one provides；

Fig. 2 A is the first schematic flow sheet that original video is divided into multiple video segment according to the Role Information in original video that the embodiment of the present invention one provides；

Fig. 2 B is the second schematic flow sheet that original video is divided into multiple video segment according to the Role Information in original video that the embodiment of the present invention one provides；

Fig. 2 C is the third schematic flow sheet that original video is divided into multiple video segment according to the Role Information in original video that the embodiment of the present invention one provides；

Fig. 2 D is the 4th kind of schematic flow sheet that original video is divided into multiple video segment according to the Role Information in original video that the embodiment of the present invention one provides；

Fig. 2 E is the 5th kind of schematic flow sheet that original video is divided into multiple video segment according to the Role Information in original video that the embodiment of the present invention one provides；

Fig. 2 F is the 6th kind of schematic flow sheet that original video is divided into multiple video segment according to the Role Information in original video that the embodiment of the present invention one provides；

Fig. 3 A is a kind of schematic flow sheet that the video segment corresponding according to target roles that the embodiment of the present invention one provides generates the search video corresponding to target roles；

Fig. 3 B is the another kind of schematic flow sheet that the video segment corresponding according to target roles that the embodiment of the present invention one provides generates the search video corresponding to target roles；

Fig. 4 A is the structural representation of a kind of video searching apparatus that the embodiment of the present invention two provides；

Fig. 4 B is the structural representation of the another kind of video searching apparatus that the embodiment of the present invention two provides；

Fig. 5 is the structural representation of the video searching apparatus that the embodiment of the present invention three provides；

Detailed description of the invention

For making the purpose of the embodiment of the present invention, technical scheme and advantage clearer, below in conjunction with the accompanying drawing in the embodiment of the present invention, technical scheme in the embodiment of the present invention is clearly and completely described, obviously, described embodiment is a part of embodiment of the present invention rather than whole embodiments.Based on the embodiment in the present invention, the every other embodiment that those of ordinary skill in the art are obtained under not making creative work premise, broadly fall into the scope of protection of the invention.

Figure 1A is the schematic flow sheet of a kind of video searching method that the embodiment of the present invention one provides.The executive agent of the method can be smart mobile phone, intelligent television, high definition set top box, panel computer, notebook computer, Ultra-Mobile PC's (English: Ultra-mobile Personal Computer, be called for short: UMPC), net book, personal digital assistant (English: Personal Digital Assistant, PDA) be called for short: the terminal such as；Meanwhile, the executive agent of the method can also be to have the application software of video search function, such as Tengxun's video, likes strange skill video, Baidu etc..As shown in Figure 1A, the video searching method that the present embodiment provides specifically includes following steps:

S11, reception video searching information, wherein, video searching information includes the Role Information of target roles, and Role Information is used for identifying different role.

S12, obtain, according to the Role Information of target roles, the video segment that target roles is corresponding.

S13, generate the search video corresponding to target roles according to video segment corresponding to target roles.

Example, video searching information can be the phonetic search information that user passes through the speech input device input of terminal unit, after speech input device on terminal unit receives phonetic search information, can be to the Role Information being used for identified role in phonetic search information, as the role's name in video resolves, thus know that user wants the target roles watched.Certainly, the video searching information of the embodiment of the present invention can also be inputted by other input modes of terminal unit, such as inputted by the touch screen of terminal unit, by the input through keyboard etc. being connected with terminal unit, obviously being merely illustrative of, the input mode of the video searching information not representing the embodiment of the present invention is confined to this herein.

Example, when the Role Information in video searching information is resolved, the different modes such as keyword match can be utilized, and retrieve in the Role Information storehouse pre-build and inquire about, judging whether to include in video searching information the Role Information of target roles, this is not limited by the present invention.

When carrying out video search, firstly the need of receiving video searching information, and video searching information includes the Role Information of target roles needed for user, this Role Information can be used for identifying different roles, the most just can carry out follow-up video segment search procedure.Concrete, because can there are different roles in video, each role occurs in different time sections in video.So, it is desirable to when obtaining the video segment of target roles, can be by the Role Information of target roles, get the video segment corresponding to this target roles, in each video segment, the most only comprise picture and the sound of this particular persons role, can be so that follow-up synthesis provides basis.

Concrete, because generally role is the specific personage of institute or other roles in a certain original video, thus according to constantly substituting the different role occurred in video, video can be divided into multiple video segment.Figure 1B is the schematic flow sheet of the another kind of video searching method that the embodiment of the present invention one provides.As shown in Figure 1B, before receiving video searching information, this video searching method can also include:

The Role Information of each role in S14, acquisition original video；

S15, according to the Role Information of each role in original video, original video is divided into multiple video segment, the Role Information of the corresponding role of each video segment.

As such, it is possible to the Role Information of each different role in acquisition original video, so that different role to be made a distinction, and according to different role, original video is carried out dividing processing.

As a example by one section of original video is carried out dividing processing, this section of original video can include tri-dominant roles of A, B and C, different according to the role of display in video, video can be divided into 20 sections, as shown in table 1.

Table 1 original video is according to the segmentation signal table of Partition of role

Wherein, whole section of original video is divided in 1-20 section, and every section of video a corresponding role or does not has role.The most corresponding role A of fragment 1,3,7,10,12,16 and 20 in original video；Fragment 4,9,11,13,15 role B corresponding with 18；Fragment 2,5,8,14,17 correspondences are role C, and fragment 6 and fragment 19 are background fragment, do not have the picture of character or sound to occur.

After original video divides, the information such as broadcasting label can be utilized, it is marked on time shaft position corresponding in original video, so when follow-up lookup, have only to obtain and specifically play label, the time shaft that this broadcasting label is corresponding on original video can be searched, thus find the video segment in original video.Example, the label of playing of the embodiment of the present invention is target video fragment time tag corresponding on the time shaft of original video, such as film " rivers and lakes of a people " the 35th minute～the 38th minute.Wherein, original video both can be one section of single video file, it is also possible to for the collection of drama being made up of multiple videos.

Concrete, it is possible to use the Role Information in original video, carry out the mark of different role in original video.Role Information can embody the feature of different role, and specific role is branched away with other role zone.General, Role Information can be text feature corresponding in video caption, it is also possible to be the role's resemblance in video frame image, or the sound characteristic etc. of role in video.

When utilizing Role Information that original video is divided, the Role Information used generally by original video is processed and role screen and get.Concrete, can in advance original video be traveled through and feature extraction, to construct a property data base, reuse mode identification or sorting technique filter out the feature of specific role.

Utilize Role Information that original video is carried out segmentation, after marking off the video segment of multiple corresponding different role, according to the Role Information of target roles, all video segments can be screened, to obtain should all video segments of target roles.Wherein, the Role Information of target roles can be inputted by user, and such as user can input role's title of target roles, the picture of role's face or one section of sound etc. of role.After obtaining the Role Information of target roles, can retrieve in video segment and screen, pick out the video segment of the Role Information meeting this target roles.General, in order to avoid omitting, need to obtain whole original video, to search out all fragments meeting this role in original video.

Or as a example by the video segmentation in table 1.Because having only to obtain the video segment corresponding to character, so negligible background fragment, and the video segment corresponding to different characters is classified, concrete division is as shown in table 2.Concrete, obtain the video segment of target roles, both can be the broadcasting label obtaining qualified video segment, it is also possible to be the editing from original video of qualified video segment to be got off, to form independent video clip files.General, in order to save resource and space, generally can only obtain the broadcasting label of target roles video segment, and be formed without new independent video clip files.

Video segment signal table corresponding to table 2 different role

Because original video is carried out Role Parsing and segmentation, need to take the substantial amounts of time and calculating process resource, in order to when reducing follow-up repeating playing take time and resource, it is also possible to saved by the video segment that editing is good.Such as, according to the Role Information in original video, after original video is divided into multiple video segment, video segment can be stored in media library.Media library can be the media library on network, it is also possible to be locally stored.In media library, store this video segment carrying out according to Role Information dividing, when the follow-up video needing and playing corresponding different role, directly can search relevant video segments from media library, and carry out follow-up splicing and synthesis.

After obtaining the video segment of the target roles intentionally got, because user is it is desired that comprise one whole section of video of this target roles content, so needing the video segment to target roles is corresponding to splice, to generate one section of complete search video corresponding to target roles.

Fig. 1 C is the schematic flow sheet of the third video searching method that the embodiment of the present invention one provides.As shown in Figure 1 C, on the basis of the embodiment shown in aforementioned Figure 1A and Figure 1B, when the video segment corresponding according to target roles generates search video corresponding to target roles, specifically may include steps of:

S131, according to video segment corresponding to time shaft sequential concatenation target roles, generate the search video that target roles is corresponding.

Because original video is when playing, all video segments all successively occur according to the order of time shaft, thus video be according to time shaft order carry out splicing obtained by, can ensure that in the video corresponding to target roles, the priority playing sequence in original segments is still conformed between different video fragment, without dislocation between video segment, disorder phenomenon occur, i.e. do not have logic or the sequencing problem of video plot, it is possible to ensure the viewing experience of user.In addition it is also possible to arrange different splicing, synthetic method, to form the video corresponding to the target roles with other effect.

In the present embodiment, when user wants to obtain the video segment corresponding to the role oneself paid close attention to, can first receive the video searching information of the Role Information including target roles, wherein, these Role Informations can be used for identifying different role, and then the Role Information further according to target roles obtains the video segment that target roles is corresponding；The video segment corresponding finally according to target roles generates the search video corresponding to target roles.The search video of correspondence so can be generated according to the Partial Fragment corresponding to specific role in video, thus allow user watch the part corresponding to the role oneself paid close attention in target video, eliminate user oneself manual search, the process of adjustment playing progress rate, the viewing being effectively increased user is experienced, and meets the individualized video search need of user.

When the video segment that acquisition specific role is corresponding, original video can be according to different roles, and it is divided into multiple video segment in advance, and the video segment good by Partition of role is stored on network or in the media library of this locality, so when needing to watch fragment corresponding to specific role, getting final product the multiple video segments corresponding to this role that directly extracting directly has been made from media library, and carry out splicing merging, synthesis is available for the video that user watches.In addition, directly original video can also be divided and process, so can when lacking existing media library, still can from original video multiple video segments corresponding to the corresponding role of extracting directly, and obtained, by these video segments, the video that this final specific role is corresponding.On the basis of previous embodiment, below for different Role Informations, provide several detailed description of the invention that original video is divided into multiple video segment respectively.

As the optional embodiment of one, when original video includes the subtitle file of correspondence, judgement and the examination of different role according to the content in subtitle file, can be carried out.Fig. 2 A is the first schematic flow sheet that original video is divided into multiple video segment according to the Role Information in original video that the embodiment of the present invention one provides.As shown in Figure 2 A, on the basis of the embodiment of aforementioned Figure 1B to Fig. 1 C, when Role Information entrained by Role Information is the subtitle file that original video is corresponding, according to the Role Information in original video, original video is divided into the step of multiple video segment, specifically can include following content:

S151, according to the Role Information entrained by subtitle file corresponding to original video subtitle file is divided into multiple text segmentation, wherein, a role in each text segmentation correspondence original video.

S152, timeline information according to each text segmentation determine the video segment that text segmentation is corresponding in original video.

Wherein, when original video includes multiple diversity video, for each diversity video.The time shaft that the subtitle file of its correspondence all can duplicate.Repeat in order to avoid the timeline information of subtitle file corresponding to different diversity videos produces, in addition it is also necessary to processing the timeline information in subtitle file, the time shaft making each diversity video correspondence subtitle file is different from the time shaft of other subtitle file.

Include the situation of multiple diversity video to adapt to original video, optionally, Fig. 2 B is the second schematic flow sheet that original video is divided into multiple video segment according to the Role Information in original video that the embodiment of the present invention one provides.As shown in Figure 2 B, on the basis of earlier figures 2A, before above-mentioned steps S151, it is also possible to comprise the steps:

S153, when original video includes multiple diversity video segment, the timeline information of multiple diversity video segment correspondence subtitle files is normalized, so that the timeline information of each subtitle file is to should uniquely show the time in original video by subtitle file.

Now, for the subtitle file that each diversity video segment is corresponding, all its timeline information can be normalized, make the timeline information of each subtitle file, all can represent this section of captions display time in whole original video uniquely, allow in original video each time period all to there being unique captions, it is to avoid because in subtitle file, timeline information is identical, and to occur distinguishing the situation of corresponding video fragment.

In the present embodiment mode, original video includes the subtitle file of correspondence, subtitle file can in imbed in original video, it is also possible to exist as plug-in unique file.Because original video includes the subtitle file of correspondence, it is possible to by Role Information entrained in subtitle file, subtitle file is carried out dividing processing.General, in the captions of original video, the captions such as personage's dialogue can include the information such as character title, now can find, by modes such as keywords, the corresponding captioned test being identified with different characters.This video segment corresponding to captioned test segmentation is the video segment of a certain role.

And when the captions of original video do not comprise role's name information, it is also possible to carried out the semantic understanding of text by natural language processing method, thus learn the main body of speaking of captioned test, and then obtain Role Information.The method of semantic understanding, can mark off roughly the captioned test segmentation corresponding to different role.

Because captions include captioned test and timeline information, so in obtaining each text segmentation of captions and original video after the corresponding relation of different role, can be by different different time axis informations corresponding to captioned test, determine the video time position that each text segmentation is corresponding in original video, and carry out the division of video segment according to this time location.

Use the Role Information entrained by subtitle file of original video, carry out the division of different role video segment, because the timeline information of subtitle file higher with the screen synchronization of original video (synchronous error of general captions and picture is within 0.1 second), during so carrying out video segment division, the scope marked off is the most accurate；And because subtitle file is usually text formatting, file size is less, it is possible to reduce the process time of video search, and reduce power consumption of processing unit.

Additionally, as alternatively possible embodiment, it is also possible to utilizing in original video, the face feature information of different role carries out the identification of role and the division of video segment.Concrete, Fig. 2 C is the third schematic flow sheet that original video is divided into multiple video segment according to the Role Information in original video that the embodiment of the present invention one provides.As shown in Figure 2 C, on the basis of Figure 1B to Fig. 1 C illustrated embodiment, when the face feature information of role during Role Information is original video, according to the Role Information in original video, original video is divided into the step of multiple video segment, specifically may include that

S154, picture frame each in original video is carried out recognition of face, to obtain the face feature information of picture frame, wherein, a role in each picture frame correspondence original video.

S155, the face feature information stored in the face feature information of each picture frame and face feature information data base is compared, with the role corresponding to the face feature information of acquisition picture frame, wherein, face feature information data base is for recording the corresponding relation between face feature information and role.

S156, according to the role corresponding to the face feature information of each picture frame, original video is divided into multiple video segment.

In present embodiment, when original video being carried out role and identifying, original video can be carried out face recognition process frame by frame.Recognition of face is a kind of biological identification technology that facial feature information based on people carries out identification, can be by picture frame in original video be detected, such as judge in picture frame either with or without the face that role occurs according to the shape description of human face and the range performance between them, after identifying face from picture frame, then the face feature information in acquired image frames.During Gai, position and the size of face is gone out firstly the need of accurate calibration, then figure and the shape facility of role's face are passed through Mathematical treatment, obtain the mathematical feature that can carry out weighing with compare, such as histogram feature, color characteristic, template characteristic, architectural feature etc..These features can be used as face feature information.Facial characteristics can represent unique role, thus is identified this role.

In acquisition original video after the face feature information of each picture frame, need to compare, so that it is determined that the face feature information on picture frame is consistent with which face feature information stored in face feature information data base the face feature information stored in the face feature information obtained and face feature information data base.Because record has the corresponding relation between face feature information and role in face feature information data base, it is possible to and then learn which character in the face feature information correspondence original video of picture frame.

All picture frames in original video are all extracted face feature information, and after carrying out the detection identification of role, can carry out the role corresponding to picture frame adding up, integrating, thus picture frame is integrated into video segment, the all corresponding independent role of each video segment, so can complete the division work of video segment.

Wherein, for carrying out the face feature information data base of face feature information comparison, can be the data base a large amount of character data being stored and being formed, can also be to sort out after original video is carried out data acquisition and processing, thus the face feature information data base obtaining this original video obtains.Because in different video, even same performer, the facial characteristics of its character moulded is also possible to there is the biggest difference, so it is the most identical in order to ensure the face feature information recorded in face feature information data base and the role's facial characteristics in original video, ensure role is had higher discrimination, generally can use and original video is carried out data process, to obtain face feature information data base based on this original video.

Carried out the identification of role by face feature information, image and the picture that can directly present according to role are identified, thus ensure that the concordance between the role and picture identified and synchronicity.

Concrete, Fig. 2 D is the 4th kind of schematic flow sheet that original video is divided into multiple video segment according to the Role Information in original video that the embodiment of the present invention one provides.As shown in Figure 2 D, on the basis of the embodiment of earlier figures 2C, in order to set up face feature information data base, when before according to the Role Information in original video original video is divided into the step of multiple video segment, it is also possible to comprise the steps:

S157, picture frame each in original video is carried out recognition of face, and gather the face feature information in the picture frame identifying face.

S158, face feature information is carried out pattern recognition, to detect the role corresponding to face feature information.

S159, the role corresponding to face feature information and face feature information is registered in face feature information data base.

Wherein, first can travel through picture frame all of in original video, each picture frame all carries out face recognition process, and the face feature information in the picture frame that may recognize that face is extracted；Then by machine learning and intelligent algorithm, face feature information being carried out pattern recognition, the face feature information that similarity exceedes certain threshold value carries out sorting out and integrated, thus obtains the role corresponding to face feature information；Finally, can face feature information be registered among face feature information data base with the role corresponding to this face feature information, in order in follow-up video segment partiting step, carry out inquiry and the comparison of face feature information.

Additionally, as alternatively possible embodiment, it is also possible to carry out the identification of role by the sound characteristic information of different role in original video.Concrete, Fig. 2 E is the 5th kind of schematic flow sheet that original video is divided into multiple video segment according to the Role Information in original video that the embodiment of the present invention one provides.As shown in Figure 2 E, on the basis of Figure 1B to Fig. 1 C illustrated embodiment, when the sound characteristic information of role during Role Information is original video, according to the Role Information in original video, original video is divided into the step of multiple video segment, specifically may include that

S1510, sound clip each in original video is carried out voice recognition, to obtain the sound characteristic information of each sound clip, wherein, a role in media sound clip correspondence original video；

S1511, the sound characteristic information stored in the sound characteristic information of each sound clip and sound characteristic information database is compared, with the role corresponding to the sound characteristic information of acquisition sound clip, wherein, sound characteristic information database is for recording the corresponding relation between sound characteristic information and role；

S1512, according to the role corresponding to the sound characteristic information of each sound clip, original video is divided into multiple video segment.

To know method for distinguishing similar with utilizing face feature information to carry out role, in present embodiment, when original video carrying out role and identifying, sound clip each in original video can be carried out voice recognition.Concrete, voice recognition can be Application on Voiceprint Recognition (Voiceprint Recognition is called for short VPR).Application on Voiceprint Recognition is also called Speaker Identification (Speaker Recognition), its basic meaning comprises two classes, i.e. speaker's identification (Speaker Identification) and speaker verification (Speaker Verification).The former is in order to judge described in which concrete role that certain section of voice comes from some people；And the latter is in order to confirm whether certain section of voice is described in the someone specified.Because Application on Voiceprint Recognition is by being identified the sound characteristic information extracting role, and unrelated with audio text, as the pronunciation content holding without the role that makes to speak, such that it is able to carry out role's identification easily.

After the sound characteristic information obtaining each sound clip, need to compare, to learn sound characteristic information correspond to which character in original video the sound characteristic information stored in the sound characteristic information obtained and sound characteristic information database.Thereafter, can carry out dividing, integrating by the video section corresponding to different sound clips, thus original video is divided into multiple video segments of corresponding different role, its concrete grammar is similar with the method shown in earlier figures with step, and here is omitted.

Carrying out the identification of role and the division of video segment by the sound characteristic information of role, because sound characteristic information is to be obtained by the audio frequency in extraction video, its data volume is less, it is possible to while accurately identifying, and reduces the process time to video.

Same, sound characteristic information database can also be set up according to the voice data in original video.Concrete, Fig. 2 F is the 6th kind of schematic flow sheet that original video is divided into multiple video segment according to the Role Information in original video that the embodiment of the present invention one provides.As shown in Figure 2 F, on the basis of the embodiment of earlier figures 2E, in order to set up sound characteristic information database, when before according to the Role Information in original video original video is divided into the step of multiple video segment, also comprise the steps:

S1513, sound clip each in original video is carried out voice recognition, and gather the sound characteristic information in the sound clip identifying voice；

S1514, sound characteristic information is carried out pattern recognition, to detect the role corresponding to sound characteristic information；

S1515, the role corresponding to sound characteristic information and sound characteristic information is registered in sound characteristic information database.

Above-mentioned concrete steps are similar with the embodiment shown in Fig. 2 E, it is all that the sound clip in original video is traveled through, to obtain the sound characteristic information having in the sound clip of voice, the methods such as recycling machine learning carry out pattern recognition to sound characteristic information, thus obtain each role corresponding to sound characteristic information, and by corresponding relation record among sound characteristic information database.And wherein many characteristics based on sound characteristic information such as vocal prints, when sound characteristic information is carried out pattern recognition, specifically can use following methods: according to hidden Markov model method and vector quantization clustering method, sound characteristic information is carried out pattern recognition, to detect the role corresponding to sound characteristic information.Using the mode that hidden Markov model method and vector quantization clustering method combine, on the one hand can obtain preferable recognition effect, on the other hand algorithm complex is the highest, it is possible to ensure processing speed, reduces processor burden.

The most each embodiment, it is the role characteristic information such as the sound characteristic to the captions in original video, the facial characteristics of role or role to be identified, to extract in original video should the corresponding video fragment of role, so that in the case of lacking media library, carry out search and the searching work of video segment, and ensure that follow-up video segment is searched, spliced and the smooth realization of synthesis step.

In addition, according to the ready-made video segment provided in media library, or according to when carrying out on the basis of the video segment that aforementioned each embodiment obtains splicing and synthesizing, when user wants dialogue, the interaction etc. of seeing between different role to coordinate performance, target roles now includes at least two role.And if only individually search according to the Role Information of the most each role, then can only obtain each self-corresponding video segment of different role, and the cooperation performance fragment between two roles cannot be screened out.In order to find out the cooperation performance fragment between two roles exactly, need comprehensively to account for two different role.Concrete, on the basis of the embodiment shown in earlier figures 1C, when target roles includes two different roles, during such as first object role and the second target roles, Role Information according to target roles obtains the step of video segment corresponding to target roles, it is specifically as follows: obtain, according to the Role Information of first object role, the video segment that first object role is corresponding, and obtain, according to the Role Information of the second target roles, the video segment that the second target roles is corresponding.

Because two roles of the performance that cooperates that first object role and the second target roles are users to be wanted to see, so when obtaining the video segment corresponding to target roles, need the Role Information each according to two different role, obtain the video segment that first object role is corresponding respectively, and second video segment corresponding to target roles, in order to carry out subsequent treatment.

On the basis of getting the video segment corresponding to two target roles, by judging video segment corresponding to two target roles relative position on a timeline and relation, thus first object role and the second target roles overall video segment when engaging in the dialogue or other coordinates performance can be extracted.

On the basis of above-mentioned embodiment, corresponding, the step generating the search video corresponding to target roles according to the video segment that target roles is corresponding is also required to be adjusted correspondingly.Fig. 3 A is a kind of schematic flow sheet that the video segment corresponding according to target roles that the embodiment of the present invention one provides generates the search video corresponding to target roles.As shown in Figure 3A, on the basis of the embodiment of earlier figures 1C, when target roles includes two different roles, during such as first object role and the second target roles, generating the search video corresponding to target roles according to the video segment that target roles is corresponding, it specifically may include that

S132, from the video segment that the second target roles is corresponding, obtain and be positioned at the video segment that the second target roles before or after the video segment that first object role is corresponding is corresponding on time shaft；

S133, according to time shaft order, the video segment that the second target roles before or after the video segment that video segment corresponding to splicing first object role is corresponding with being positioned at first object role is corresponding, the search video that generation first object role is corresponding.

According to said method, when the video segment that video segment corresponding to first object role and the second target roles are corresponding being spliced according to time shaft order, firstly for the video segment corresponding to each first object role, all obtain and be positioned at the video segment that the second target roles before or after this video segment is corresponding on a timeline.With the second corresponding video segment of target roles being positioned at before or after this video segment, can be the previous video segment of video segment corresponding to this first object role or a rear video segment, can also be the video segment between video segment corresponding with this first object role with certain intervals, if the time interval between two video segments is less than certain time threshold value etc..As first looked for the video segment corresponding to one of them role, then need to judge in original video whether the video segment before or after this video segment is the video segment corresponding to another role, if, then explanation has found the video segment coordinating performance between first object role with the second target roles, by the two video segment that finds together as the video segment corresponding to target roles；And if the video segment corresponding to one of them role found is not the video segment corresponding to another role, but other role's or background video fragment, then in explanation the two video segment, do not carry out coordinating performance between first object role with the second target roles, therefore the two video segment should be cast out.Wherein, with the second corresponding video segment of target roles being positioned at before or after this video segment, can be the previous video segment of video segment corresponding to this first object role or a rear video segment, can also be the video segment between video segment corresponding with this first object role with certain intervals, if the time interval between two video segments is less than certain time threshold value etc..

After the video segment that the second target roles before or after finding the first video segment meeting condition and being positioned at the video segment that this first object role is corresponding is corresponding, these video segments can be stitched together, generate the video segment corresponding to first object role.

According to said method, all of video segment can be made a look up judgement, thus obtain the video segment that all of target roles is corresponding, and the splicing of these video segments is gathered, thus generate first object role and coordinate video corresponding when performing with the second target roles.If target roles has multiple, similar when also with two roles of its search strategy coordinate performance, here is omitted.

Pass through said method, it is possible to obtain coordinate video segment when performing engage in the dialogue, interactive etc. between different role, allow user when watching the role of needs, additionally it is possible to obtain the more perfect story of a play or opera.

Further, when searching the video segment that first object role is corresponding with the second target roles, and when carrying out splicing or the synthesis of video segment, between video segment and target roles accessed by guarantee, there is stronger relatedness, it is also possible to retrain further when obtaining the second target roles correspondence video segment.Fig. 3 B is the another kind of schematic flow sheet that the video segment corresponding according to target roles that the embodiment of the present invention one provides generates the search video corresponding to target roles.As shown in Figure 3 B, on the basis of earlier figures 1C, generate the step of the search video corresponding to target roles according to the video segment that target roles is corresponding, specifically comprise the steps that

S134, from the video segment that the second target roles is corresponding, obtain on a timeline, a corresponding upper video segment or next video segment are the video segment that the second target roles of video segment corresponding to first object role is corresponding；

S135, according to time shaft order, splicing video segment corresponding to first object role and a upper video segment corresponding on a timeline or next video segment are the video segment that the second target roles of video segment corresponding to first object role is corresponding, the search video that generation first object role is corresponding.

According to above-mentioned steps, when obtaining the second target roles correspondence video segment meeting condition, may determine that whether video segment corresponding to this second target roles is connected with first object role's correspondence video segment, it is first object role's correspondence video segment upper video segment on a timeline or next video segment, if, the video segment that then this second target roles is corresponding meets requirement exactly, and the second corresponding video segment of target roles of splicing synthesis can be carried out with first object role, if it is not, then this second target roles is rejected.

Go out the second target roles correspondence video segment of all satisfied requirements according to above-mentioned conditional filtering after, can be according to time shaft order, the video segment that the video segment that these the second target roles are corresponding is corresponding with first object role is spliced, thus generates the search video that first object role is corresponding.

In addition, because target roles may there are interactive scene with other roles, or compartment of terrain occurs in the plot that one section of relatedness is stronger, now, the video segment of this target roles occurs iff selection, and given up the video segment having stronger plot to associate with this target roles, and the video being spliced will be made to seem incoherent, the viewing greatly reducing user is experienced.In order to avoid above-mentioned phenomenon, when searching the video segment corresponding to target roles, it is also possible to obtain and the related plot of target roles or scene.Concrete, according to the video segment that time shaft sequential concatenation target roles is corresponding, generate search video corresponding to target roles and can comprise the following steps that the video segment that there is other role between the first video segment and the second video segment found, and the time shaft interval that first between video segment and the second video segment less than preset time threshold time, by the first video segment, second video segment and the video segment between the first video segment with the second video segment are together as video segment corresponding to target roles, wherein, first video segment and the second video segment are the video segment that target roles is corresponding, and first video segment and the second video segment countershaft on time order arrangement.

Wherein, the first video segment and the second video segment are the video segment that found target roles is corresponding, and the first video segment and the second video segment are according to time shaft order in tandem.Time shaft interval between the first video segment and the second video segment is less than certain value, during such as preset time threshold, then illustrates to be likely that there are plot association between the first video segment and the second video segment, or is under Same Scene.And now in order to ensure the fluency of plot, need to extract video segment all of between the first video segment and the second video segment together with above-mentioned two video segment, with collectively as the video segment corresponding to target roles.In this section of video segment, not only include the video segment corresponding to target roles, other role or the video segment such as action, scene are may also contain, these video segments together form the plot that under a certain scene of this target roles is complete, coherent, it is thus possible to effectively ensures the viewing effect of user.

It should be noted that whether there is the preset time threshold of plot or scene relating for differentiating the first video segment and the second video segment, could be arranged to different values, such as 5 10 seconds etc..In addition, different value can also be set according to the dissimilar of original video, such as rhythm action movie faster, then can be arranged by preset time threshold is shorter, and during the type such as original video romance movie that to be rhythm slower, longer preset time threshold can be set, so can improve accuracy corresponding between video segment with target roles further.

According to said method, it is possible to obtain the overall fragment of the plot relevant with target roles and scene, it is to avoid the video being spliced is the most scrappy, and has influence on the problem that user's viewing is experienced, and makes user can enjoy the most complete plot fragment.

Fig. 4 A is the structural representation of a kind of video searching apparatus that the embodiment of the present invention two provides.The video searching apparatus that the present embodiment provides can perform the video searching method described in previous embodiment one.Concrete, as shown in Figure 4 A, the video searching apparatus 200 that the present embodiment provides specifically includes:

Receiver module 21, is used for receiving video searching information, and wherein, video searching information includes the Role Information of target roles, and Role Information is used for identifying different role；

Video acquiring module 22, obtains, for the Role Information according to target roles, the video segment that target roles is corresponding；

Video-splicing module 23, generates the search video corresponding to target roles for the video segment corresponding according to target roles.

Video searching apparatus is before the video segment obtaining different role, it is necessary first to receives video searching information, and finds, by the Role Information of target roles included in video searching information, all video segments that this target roles is corresponding.These video segments can be the most divided existing fragment completed, it is also possible to is first to get original video, carries out further according to original video dividing.

Because generally role is the specific personage of institute or other roles in a certain original video, thus according to constantly substituting the different role occurred in video, video can be divided into multiple video segment.Fig. 4 B is the structural representation of the another kind of video searching apparatus that the embodiment of the present invention two provides.As shown in Figure 4 B, in order to original video being divided into multiple fragment according to different role, video searching apparatus also includes role's processing module 24 and video processing module 25, role's processing module 24 is for before receiving video searching information, the Role Information of each role in acquisition original video, original video, for according to the Role Information of each role in original video, is divided into multiple video segment by video processing module 25, the Role Information of the corresponding role of each video segment.

When original video is carried out video search, need according to the Role Information in original video, original video is divided into multiple video segment.Concrete, in original video, there are different roles, each role occurs in the different time sections in original video.So, according to constantly substituting the different role occurred in video, video can be divided into multiple video segment, in each video segment, the most only comprise picture and the sound of a character, so original video can be carried out subregion according to difference role occur or piecemeal processes, in order to follow-up selects and synthesize.

Utilize Role Information that original video is carried out segmentation, after marking off the video segment of multiple corresponding different role, according to the Role Information of target roles, all video segments can be screened, to obtain should the video segment of target roles.Wherein, the Role Information of target roles can be inputted by user, and such as user can input role's title of target roles, the picture of role's face or one section of sound etc. of role.After obtaining the Role Information of target roles, can retrieve in video segment and screen, pick out the video segment of the Role Information meeting this target roles.

Optionally, as a kind of enforceable mode, Role Information is the Role Information entrained by the subtitle file that original video is corresponding,

Video processing module 25 specifically for:

According to the Role Information entrained by the subtitle file that original video is corresponding, subtitle file is divided into multiple text segmentation, wherein, a role in each text segmentation correspondence original video；

Timeline information according to each text segmentation determines the video segment that text segmentation is corresponding in original video.

When original video is made up of multiple diversity videos, optionally, before subtitle file being divided into multiple text segmentation according to the Role Information entrained by the subtitle file that original video is corresponding, video processing module 25 is additionally operable to:

When original video includes multiple diversity video segment, the timeline information of multiple diversity video segment correspondence subtitle files is normalized, so that the timeline information of each subtitle file is to should uniquely show the time in original video by subtitle file.

Optionally, as another kind of enforceable mode, Role Information is the face feature information of role in original video, role's processing module 24 specifically for:

Picture frame each in original video is carried out recognition of face, to obtain the face feature information of picture frame, wherein, a role in each picture frame correspondence original video；

The face feature information stored in the face feature information of each picture frame and face feature information data base is compared, with the role corresponding to the face feature information of acquisition picture frame, wherein, face feature information data base is for recording the corresponding relation between face feature information and role；

Video processing module 25 is additionally operable to: according to the role corresponding to the face feature information of each picture frame, and original video is divided into multiple video segment.

When original video being carried out role and identifying, original video can be carried out face recognition process frame by frame.Facial characteristics can represent unique role, thus is identified this role.In acquisition original video after the face feature information of each picture frame, need to compare, so that it is determined that the face feature information on picture frame is consistent with which face feature information stored in face feature information data base the face feature information stored in the face feature information obtained and face feature information data base.Because record has the corresponding relation between face feature information and role in face feature information data base, it is possible to and then learn which character in the face feature information correspondence original video of picture frame.

Optionally, on the basis of above-mentioned embodiment, before original video being divided into multiple video segment according to the Role Information in original video, role's processing module 24 can be also used for:

Picture frame each in original video is carried out recognition of face, and gathers the face feature information in the picture frame identifying face；

Face feature information is carried out pattern recognition, to detect the role corresponding to face feature information；

Face feature information is registered in face feature information data base with the role corresponding to face feature information.

Optionally, as another kind of enforceable mode, Role Information is the sound characteristic information of role in original video, now, role's processing module 24 specifically for:

Sound clip each in original video is carried out voice recognition, to obtain the sound characteristic information of each sound clip, wherein, a role in media sound clip correspondence original video；

The sound characteristic information stored in the sound characteristic information of each sound clip and sound characteristic information database is compared, with the role corresponding to the sound characteristic information of acquisition sound clip, wherein, sound characteristic information database is for recording the corresponding relation between sound characteristic information and role；

Original video, for according to the role corresponding to the sound characteristic information of each sound clip, is divided into multiple video segment by video processing module 25.

After the sound characteristic information obtaining each sound clip, need to compare, to learn sound characteristic information correspond to which character in original video the sound characteristic information stored in the sound characteristic information obtained and sound characteristic information database.Thereafter, can carry out dividing, integrating by the video section corresponding to different sound clips, thus original video is divided into multiple video segments of corresponding different role.

Carrying out the identification of role and the division of video segment by the sound characteristic information of role, because sound characteristic information is to be obtained by the audio frequency in extraction video, its data volume is less, it is possible to while accurately identifying, and reduces the process time to video.Optionally, on the basis of a upper embodiment, before original video being divided into multiple video segment according to the Role Information in original video, role's processing module 24 can be also used for:

Sound clip each in original video is carried out voice recognition, and gathers the sound characteristic information in the sound clip identifying voice；

Sound characteristic information is carried out pattern recognition, to detect the role corresponding to sound characteristic information；

Sound characteristic information is registered in sound characteristic information database with the role corresponding to sound characteristic information.

Wherein, optionally, role's processing module 24 specifically can be also used for:

According to hidden Markov model method and vector quantization clustering method, sound characteristic information is carried out pattern recognition, to detect the role corresponding to sound characteristic information.

After having obtained the video segment that target roles is corresponding, because being respectively provided with regular hour sequencing between video segment, so after obtaining the video segment that target roles is corresponding, video-splicing module 23 specifically for:

According to the video segment that time shaft sequential concatenation target roles is corresponding, generate the search video that target roles is corresponding.

Optionally, as another kind of enforceable mode, video-splicing module 23 specifically may be used for:

According to all video segments that time shaft sequential concatenation target roles is corresponding, and as the video corresponding to target roles.

According to time shaft order, video segment is spliced, it is ensured that target roles correspondence video there is succession and logicality in time, it is ensured that the viewing of user is experienced.

Optionally, as another kind of enforceable mode, target roles includes first object role and the second target roles, video acquiring module 22 specifically for:

Role Information according to first object role obtains the video segment that first object role is corresponding, and obtains, according to the Role Information of the second target roles, the video segment that the second target roles is corresponding.

Optionally, as another kind of enforceable mode, video-splicing module 23 includes:

Determine submodule 231, for from the video segment that the second target roles is corresponding, determine the video segment that the second target roles that a upper video segment corresponding on a timeline or next video segment are video segment corresponding to first object role is corresponding；

Splicing submodule 232, for according to time shaft order, splicing video segment corresponding to first object role and a upper video segment corresponding on a timeline or next video segment are the video segment that the second target roles of video segment corresponding to first object role is corresponding, the search video that generation first object role is corresponding.

Additionally, video acquiring module 22 can be also used for:

The video segment of other role is there is between the first video segment found and the second video segment, and the time shaft interval that first between video segment and the second video segment less than preset time threshold time, by the first video segment, the second video segment and all video segments between the first video segment with the second video segment together as video segment corresponding to target roles, wherein, first video segment and the second video segment are the video segment that target roles is corresponding, and the first video segment and the second video segment countershaft on time order arrange.

Technical scheme based on above-described embodiment, according to the Role Information of role, original video can be divided, obtain being belonging respectively to the video segment of different role, user is when viewing, can be by inputting specific role, and search and watch in video the Partial Fragment corresponding to this role, user's degree of freedom when watching video is higher, has more preferably viewing and experiences.

In the present embodiment, the demand of the video segment corresponding to the role oneself paid close attention to is wished to obtain in order to meet user, video searching apparatus may particularly include the receiver module for receiving video searching information, wherein, video searching information includes the Role Information of target roles, Role Information is used for identifying different role, the video acquiring module of video segment corresponding to target roles is obtained for the Role Information according to target roles, and the video-splicing module for the search video corresponding to the video segment generation target roles corresponding according to target roles, and the corresponding role of the most each video segment, Role Information is for identifying the different role in original video.User so can be allowed to watch the Partial Fragment corresponding to the role oneself paid close attention in video, eliminate user oneself manual search, the process of adjustment playing progress rate, the viewing being effectively increased user is experienced, and meets the individualized video search need of user.

Fig. 5 is the structural representation of the video searching apparatus that the embodiment of the present invention three provides.Video searching apparatus in the present embodiment for the video searching method performing in previous embodiment one, concrete handling process, identical with the handling process of said method, be not repeated in this embodiment of the present invention.As it is shown in figure 5, the video searching apparatus that the present embodiment provides specifically includes:

Receptor 31, is used for receiving video searching information, and wherein, this video searching information includes the Role Information of target roles, and this Role Information is used for identifying different role；

Memorizer 32, is used for storing program；Specifically, program can include program code, and program code includes computer-managed instruction.

Processor 33, for performing the program that memorizer 32 is stored, obtains, with the Role Information according to target roles, the video segment that target roles is corresponding；And generate the search video corresponding to target roles according to the video segment that target roles is corresponding.

Additionally, optional, video searching apparatus also includes:

Video acquisition interface 34, is used for obtaining original video；

Processor 33 is additionally operable to: according to the Role Information of each role in original video, and original video is divided into multiple video segment, the Role Information of the corresponding role of each video segment.

Wherein, receptor 31 and video acquisition interface 34 are attached with intelligent television or other playback terminals, are used for obtaining and export various data and instruction.Memorizer 32 can comprise various RAM memory or nonvolatile memory (non-volatile memory).And the form of processor 33 may be central processing unit (Central Processing Unit, referred to as CPU), or specific integrated circuit (Application Specific Integrated Circuit, referred to as ASIC), then or it is configured to implement one or more integrated circuits of the embodiment of the present invention.Processor 33 is the control centre of video searching apparatus, utilize various interface and the various piece of the whole device of connection, it is stored in the software program in memorizer 32 and/or module by running or performing, and call the data being stored in memorizer 32, perform the various functions of device and process data, thus realizing video search function.

On implementing, realize if receptor 31, video acquisition interface 34, processor 33 and memorizer 32 are independent, then receptor 31, video acquisition interface 34, processor 33 and memorizer 32 can be connected with each other and complete mutual communicating by bus.Described bus can be industry standard architecture (Industry Standard Architecture, referred to as ISA) bus, external equipment interconnection (Peripheral Component, referred to as PCI) bus or extended industry-standard architecture (Extended Industry Standard Architecture, referred to as EISA) bus etc..Bus can be divided into address bus, data/address bus, control bus etc..For ease of representing, figure only represents with a thick line, it is not intended that an only bus or a type of bus.

In the present embodiment, video searching apparatus can first pass through receptor and receive video searching information, wherein, includes the Role Information of target roles in video searching information, and Role Information is used for identifying different role；Read the program in memorizer by processor again, obtain, with the Role Information according to target roles, the video segment that target roles is corresponding；The video segment corresponding finally according to target roles generates the search video corresponding to target roles.User so can be allowed to watch the Partial Fragment corresponding to the role oneself paid close attention in video, eliminate user oneself manual search, the process of adjustment playing progress rate, the viewing being effectively increased user is experienced, and meets the individualized video search need of user.

Last it is noted that various embodiments above is only in order to illustrate technical scheme, it is not intended to limit；Although the present invention being described in detail with reference to foregoing embodiments, it will be understood by those within the art that: the technical scheme described in foregoing embodiments still can be modified by it, or the most some or all of technical characteristic is carried out equivalent；And these amendments or replacement, do not make the essence of appropriate technical solution depart from the scope of various embodiments of the present invention technical scheme.

Claims

1. a video searching method, it is characterised in that including:

Receiving video searching information, wherein, described video searching information includes role's letter of target roles Breath, described Role Information is used for identifying different role；

Role Information according to described target roles obtains the video segment that described target roles is corresponding；

The search video that described target roles is corresponding is generated according to the video segment that described target roles is corresponding.

Video searching method the most according to claim 1, it is characterised in that described reception video is searched Before rope information, also include:

Obtain the Role Information of each role in original video；

According to the Role Information of each described role in described original video, described original video is divided into Multiple video segments, the Role Information of the corresponding described role of each described video segment.

3. according to the video searching method described in any one of claim 1-2, it is characterised in that described The video segment corresponding according to described target roles generates the search video that described target roles is corresponding, specifically wraps Include:

According to the video segment that target roles described in time shaft sequential concatenation is corresponding, generate described target roles Corresponding search video.

Video searching method the most according to claim 3, it is characterised in that described target roles bag Including first object role and the second target roles, the described Role Information according to described target roles obtains institute State the video segment that target roles is corresponding, specifically include:

Role Information according to described first object role obtains the piece of video that described first object role is corresponding Section, and obtain, according to the Role Information of described second target roles, the video that described second target roles is corresponding Fragment.

Video searching method the most according to claim 4, it is characterised in that described according to time shaft The video segment that target roles described in sequential concatenation is corresponding, generates the search video that described target roles is corresponding, Specifically include:

From the video segment that described second target roles is corresponding, obtain and be positioned at described first mesh on time shaft Mark the video segment that described second target roles before or after the video segment that role is corresponding is corresponding；

According to time shaft order, splice video segment corresponding to described first object role and be positioned at described the The piece of video that described second target roles before or after the video segment that one target roles is corresponding is corresponding Section, generates the search video that described first object role is corresponding.

From the video segment that described second target roles is corresponding, obtain on a timeline, corresponding upper one Video segment or next video segment are described second mesh of video segment corresponding to described first object role The video segment that mark role is corresponding；

According to time shaft order, splice video segment corresponding to described first object role and on a timeline A corresponding upper video segment or next video segment are the video segment that described first object role is corresponding The video segment that described second target roles is corresponding, generates the search video that described first object role is corresponding.

7. a video searching apparatus, it is characterised in that including:

Receiver module, is used for receiving video searching information, and wherein, described video searching information includes mesh The Role Information of mark role, described Role Information is used for identifying different role；

Video acquiring module, obtains described target roles for the Role Information according to target roles corresponding Video segment；

Video-splicing module, generates described target angle for the video segment corresponding according to described target roles The search video that color is corresponding.

Video searching apparatus the most according to claim 7, it is characterised in that described video is searched Rope device also includes:

Role's acquisition module, for obtaining the Role Information of each role in original video；

Video processing module, is used for according to the Role Information of each described role in described original video, Described original video is divided into multiple video segment, the corresponding described angle of each described video segment The Role Information of color.

9. according to the video searching apparatus described in any one of claim 7-8, it is characterised in that described Video-splicing module specifically for:

Video searching apparatus the most according to claim 9, described target roles includes the first mesh Mark role and the second target roles, it is characterised in that described video acquiring module specifically for:

11. video searching apparatus according to claim 10, it is characterised in that described video Concatenation module includes:

Determine submodule, for from the video segment that described second target roles is corresponding, determine time A upper video segment corresponding on countershaft or next video segment are the video that first object role is corresponding The video segment that described second target roles of fragment is corresponding；

Splicing submodule, for according to time shaft order, splices the video that described first object role is corresponding Fragment and a upper video segment corresponding on a timeline or next video segment are that first object role is corresponding Video segment corresponding to described second target roles of video segment, generate described first object role couple The search video answered.

12. 1 kinds of video searching apparatus, it is characterised in that including:

Receptor, is used for receiving video searching information, and wherein, described video searching information includes target The Role Information of role, described Role Information is used for identifying different role；

Memorizer, is used for storing program；

Processor, for performing the program of described memorizer storage, with the Role Information according to target roles Obtain the video segment that described target roles is corresponding, and raw according to the video segment that described target roles is corresponding Become the search video that described target roles is corresponding.

13. video searching apparatus according to claim 12, it is characterised in that also include:

Video acquisition interface, is used for obtaining original video；

Described processor is additionally operable to: according to the Role Information of each described role in described original video, will Described original video is divided into multiple video segment, the corresponding described role's of each described video segment Role Information.