KR20000054561A - A network-based video data retrieving system using a video indexing formula and operating method thereof - Google Patents

Abstract

PURPOSE: A network based video searching system using a video indexing system and the operating method are provided so that users can watch their concerning scenes selectively for all the programs of internet broadcasting. CONSTITUTION: A video information searching service server(300) detects scenes by fields from motion image information, which will be transmitted to the external through a network communications network(100), according to a preset basis automatically, segments them and generates the search information of representative stationary images. A user terminal(200) carries out communication connection to the video information searching service server(300) through the network communications network(100) and outputs a main webpage provided from the video information searching service server(300) on the screen. The user terminal(200) receives a webpage for motion image information output or a webpage having motion image search information provided from the video information searching service server(300) according to a user request signal inputted through the main webpage and outputs it on the screen.

Description

Network-based video retrieval system using video indexing and its operation method {A NETWORK-BASED VIDEO DATA RETRIEVING SYSTEM USING A VIDEO INDEXING FORMULA AND OPERATING METHOD THEREOF}

The present invention relates to a network-based video retrieval system using a video indexing scheme and a method of operating the same.

More specifically, a video for generating scene or content search information corresponding to a video of all programs broadcast through the Internet so that the user can select and view only scenes of interest for all programs of the Internet broadcast. A network based video retrieval system using an indexing method and a method of operating the same.

The existing video information service provided by the Internet on the Internet by conventional broadcasting stations such as MBC, KBS, SBS, etc. is being provided in the program unit of each broadcasting station. In other words, the conventional Internet broadcast viewer has a problem that it takes a lot of time and effort because they have to search sequentially to select and view the scene of the content that they want to watch from a predetermined program.

In particular, programs such as sports, news, or soap operas are already broadcast and know the whole contents, so it is common to see only the contents of news of interest, highlights of sports, or any part of soap operas, rather than the whole program.

Accordingly, the necessity for providing Internet broadcasting system viewers with the content index for each scene is provided as a representative image by dividing the contents of the program by scenes.

An object of the present invention is to separately create scene-specific or content-specific search information corresponding to the video of all programs broadcast over the Internet to solve the above-mentioned problems, and scenes according to the user's selection to watch the broadcast program over the Internet Provides a web page with search information by stars or contents as a representative image separately from a web page that can be broadcasted through the Internet or a video page that shows broadcasted video information. The present invention provides a network-based video retrieval system using a video indexing method and a method of operating the same.

1 is a block diagram illustrating a configuration of a network-based video scene retrieval system using a video indexing method according to the present invention;

FIG. 2 is a block diagram showing the configuration of a video information retrieval service server applied to FIG. 1;

3 is a flowchart illustrating a method of operating a network-based video retrieval system using a video indexing scheme according to the present invention;

4 is a block diagram for explaining a video scene automatic segmentation method applied to the present invention;

5 is an operation flowchart for explaining a video scene automatic segmentation method applied to the present invention;

6 to 9 are flowcharts illustrating the subroutine applied to FIG. 5 in more detail.

10 is an operation flowchart for explaining a caption analysis method applied to the present invention;

11 is a flowchart illustrating a scene segmentation method applied to the present invention in a more specific embodiment;

12 to 17 are operational flowcharts for describing in more detail the subroutine applied to FIG.

18 is an operation flowchart for explaining a subroutine applied to FIG. 3;

19 is a view for explaining a web page for a user interface for the convenience of a user;

20 to 23 are screen diagrams for explaining an embodiment of a web page for news search generated by the present invention.

* Description of Signs of Main Parts of Drawings *

100: network communication network 200: user terminal

300: video information retrieval service server

310: video input unit 320: video analysis unit

330: Service Department

In order to achieve the above object, the system of the present invention automatically detects and divides a scene according to a predetermined criterion for each field of video information to be transmitted to the outside through a network to generate search information, and communicates through a network communication network. When the access request is input, the main web page is transmitted, and in response to the user's request signal input through the main web page, the user can watch a video or broadcast by providing a web page for video output or a web page having video search information. A video information retrieval service server for extracting and viewing a desired scene from the recorded video, and a main web provided by the video information retrieval service server that performs a communication connection to the video information retrieval service server through a network communication network. Print pages on screen According to the user's request signal input from the web page includes a user terminal for receiving the output received on the screen, the web page having a web page or video information for output video search information supplied from the video search service server.

In addition, the method of the present invention for achieving the above object, (1) the video information retrieval service server checks the field of the video information to be transmitted to the user terminal through a network communication network, and in accordance with a predetermined condition corresponding to the checked field (2) a user terminal executing a communication program according to a user's operation to establish a communication connection to a video information retrieval service server; and (3) video information. The search service server transmits a main web page for receiving video information or search information to the user terminal according to a user's communication connection request, and the user searches for predetermined video information or search through the main web page received through the user terminal. The method may further include receiving a web page having only information.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1 is a block diagram illustrating a configuration of a network-based video retrieval system using a video indexing method according to the present invention, and FIG. 2 is a block diagram illustrating a configuration of a video information retrieval service server applied to FIG. 1.

As shown in FIG. 1, the network communication network 100 includes a wired / wireless communication network, and only a scene where an unspecified user wants to see a broadcast signal broadcast through the Internet using search information consisting of predetermined keywords. A communication connection may be made to the video information retrieval service server 300 for viewing.

The video information retrieval service server 300 automatically detects and divides a scene according to a preset criteria for each field of video information to be transmitted to the outside through the network communication network 100 to generate search information of a representative still image, and the network When a communication connection request is input through the communication network 100, the main web page is transmitted to the user terminal 200, and the web page for outputting video or video search information is output in response to a user's request signal input through the main web page. It provides the user with a web page so that the user can watch a video or extract a desired scene from the broadcast video.

The user terminal 200 performs a communication connection to the above-described video information retrieval service server 300 through the network communication network 100, and displays a main web page provided by the video information retrieval service server 300 which has established a communication connection. And outputs a web page for outputting video information or a web page having video search information supplied from the video information search service server 300 according to a user's request signal input through the main web page. .

The above-described video information retrieval service server 300 is composed of detailed components as shown in FIG. 2, which will be described in more detail as follows.

The video input unit 310 includes a subtitle decoder 311 that is included in digital broadcast data input from the outside according to the intention of the server operator to extract closed caption information, and the analog broadcast data input from the outside according to the intention of the server operator. MPEG encoder 312 for converting data and a controller 313 for checking subtitle information or outputting a control signal for MPEG data conversion by checking broadcast data inputted from the outside, which is reserved by a server operator and manufactured separately. A signal or a digital broadcast signal is received, or an analog broadcast signal or a digital broadcast signal produced by an existing broadcast station is received.

The video analyzer 320 adds a time tag generated by using a system clock to the caption information output through the caption decoder 311, and produces an index image material using the time tag added caption information as a noun unit. And a scene divider 321 for automatically generating divided video information by receiving a signal output through the MPEG encoder 312 and an analog broadcast signal, and the caption analyzer 322 or the like. The index image data generated by the scene divider 321 and the additional information input unit 323 for generating additional information according to a manipulation of a server operator and providing the same to a user, and a video output through a digital broadcast signal and an MPEG encoder. Stream file producer 324 is produced as a stream file to provide a broadcast signal through the network communication network (100).

The service unit 330 may include a service interface unit 331 for generating a web page having video search information and a web page having video search information generated through the service interface unit 331 through the network communication network 100. A plurality of stream servers 333 (1) to 333 (N) for transmitting the web server 332 to the user and the stream file produced by the stream file producer 324 to the user via the network communication network 100; Is made of.

The scene divider 321 of the video analyzer 320 receives the MPEG bit stream data and divides the video into audio and audio information, and demultiplexes and synchronizes the module 110 to produce a table of time stamps generated during the division. , Using the video image separated by the demultiplexing and synchronization module 110 to calculate the difference value for each frame between the still images, apply the difference value to the median filter, and input it to the neural network circuit to change the scene. The video segmentation module 120 is configured to detect and separate the scenes, and the silent section is calculated from the audio data separated by the demultiplexing and synchronization module to determine the silent section as a threshold for a certain period. When the threshold value is exceeded, the audio segmentation module 130 and the video segmentation module 120 recognize the start of a new frame. And the information fusion module 140 that determines the time point at which the scene changes by applying the result made by the audio splitting module 130 to the conditional probability.

The operation of the network-based video retrieval system using the video indexing scheme according to the present invention configured as described above will be described in more detail with reference to the accompanying drawings.

First, the video information retrieval service server 300 checks the field of the video information to be transmitted to the user terminal 200 through the network communication network 300, and displays the scene of the video information according to a preset condition corresponding to the checked field. A routine for generating the search information by dividing is performed (S100).

First, in order to generate the above-mentioned search information, a search information generation processor for dividing the input digital video data should be performed. The division process will be described in more detail with reference to FIGS. 4 to 19 as follows. .

As shown in FIG. 4 and FIG. 5, a method of dividing scenes of moving image data includes: splitting MPEG stream data into video stream data and audio stream data through the demultiplexing and synchronization module 110 ( In operation S110, the video segmentation module 120 detects a change point of a scene by identifying a characteristic of an image on a video screen by using the video stream data split in step S110 (S120), and the audio stream data split in step S110. The audio segmentation module 130 detects the silent section by comparing with the preset threshold value and recognizes it as a transition point of the scene (S130).

The data fusion module 140 detects the final scene change point and converts the data into data using the video stream data divided in the above-described S120 process and the audio stream data divided in the above-described S130 process (S140).

Meanwhile, the demultiplexing and synchronization process of the above-described S110 process is performed by receiving an MPEG bit stream as shown in FIG. 6 (S111), after dividing the MPEG bit data into video stream data and audio stream data (S112). In step S113, a table is added by adding time stamp information to the divided video stream data and the audio stream data.

In the method of finding a scene change point in the video segmentation process of the above-described S120 process, as shown in FIG. 7, video stream data divided through the above-described process S112 and inputted to the video segmentation module 120 is inputted. It is determined whether it is the last frame (S121).

As a result of the determination of step S121 described above, when the input video stream data is not the last frame, only the I picture region of the video stream data is extracted (S122), and the DC image of the extracted I picture is extracted (S123).

In operation S123, the video stream data having only the extracted DC image is input to calculate a difference value for each frame between the previous still image and the next still image (S124). In this case, the classification criteria are calculated based on the histogram difference of the continuous images and the difference value between pixels.

The calculated difference value for each frame is filtered through the median filter (S125), and the difference value for each frame filtered by the above-described median filter is input to a neural network having a size of 3 × 5 × 2 to perform a learning process. (S126).

As described above, if the value output through the learning process is greater than or equal to the preset value, the new scene is registered as the start of the new scene (S128). Otherwise, the process is performed again from the above-described process S121 (S127).

In addition, the audio splitting routine performed by the audio splitting module 130 receives the audio stream data divided in the above-described step S112 as shown in FIG. 8 and determines whether it is the last frame (S131).

As a result of the above-described determination in step S131, if it is not the last frame, the silent section is calculated (S132), and it is determined whether the calculated silent section exceeds a predetermined period (S133). As a result of the determination, if the calculated silent section exceeds a predetermined period of time, it is registered as the start of a new scene (S134).

As described above, when information on a predetermined scene is generated by the video stream data and the audio stream data, the information fusion module 140 receives the above-described information and detects a final scene change time point to generate search information. This will be described in more detail with reference to FIG. 9, and the video stream data analysis result of step S120 and the audio stream data analysis result of step S130 are received (S141) (S142), and it is determined whether it is the last scene change point (S143). ).

If the determination result of step S143 is not the last switching point, the analysis result of the video stream data and the audio stream data are synthesized, the synthesized data is recognized as a statistical model (S144), and then it is determined whether the scene is a switching point. In the case of the scene change point, the final scene change point is found and data-formed (S145) (S146).

Meanwhile, as illustrated in FIG. 10, the caption decoder 311 receives broadcast content in which captions are broadcast and extracts only pure text data (S410). After extracting only pure text data, a time tag is added based on a system clock existing in a computer system, and the time tag to be added is added in units of data coming into the input from the subtitle decoder 311 (S420).

That is, when data input to the subtitle decoder 311 comes in sentence units, time tags are attached in sentence units. Since the caption information thus obtained is generally not exactly synchronized with the video signal, the closed caption information is analyzed by the caption analyzer 322, and the caption information is synchronized with the analyzed caption information with reference to the video signal. Extract additional information using special characters.

The caption information thus obtained is analyzed for morphological analysis of subtitles for later retrieval to produce index material in unit of nouns (S430), and outputs subtitle information noun index information with time tags added to the additional information input unit 323 (S440). ).

The text index information thus obtained is used when searching in the news service.

As described above, video data divided by scenes by the scene divider 321 and input to the service interface unit 331 should be divided by meaningful scenes.

In this case, as shown in FIG. 11, a method of segmenting applies different analysis methods according to the contents of a video to provide a service that meets a user's needs. For this, see FIGS. 12 to 18. The description is as follows.

First, as shown in the accompanying drawings, the video contents are described first in the news video as follows.

First, news videos are divided into news chunks, which are the basic video units. Here, a news chunk refers to a video unit in which an anchor appears followed by a reporter. In order to automatically separate it, subtitle information input simultaneously with the video image of the news is used. That is, it is determined whether closed caption information input simultaneously with the news video image is input (S11), and when closed caption information is input, the positions of words such as 'anchor' and 'reporter' are extracted (S12).

After extracting the positions of words such as 'anchor' and 'reporter' in step S12 described above (S13), it is determined whether the last position is checked by checking the start position of the next scene (S14), and whether news video information is inputted. By judging, the anchor face existing in the news is automatically extracted, and the time at this point in time is extracted (S17) (S18).

On the other hand, after extracting the position of the words such as 'anchor', 'reporter' in the above-described step S12 and detects the start position of the next scene to determine whether it is the last position (S14).

As a result of the above-described determination in step S14, when the reference time is not the last position or when the anchor face is extracted through the above-described step S18, the reference of the image and the caption synchronization are set (S15). By correctly classifying the news video information into chunks and generating search information for searching the separated news chunks (S16).

Meanwhile, referring to FIG. 13, a method of extracting a screen representing the news chunk and the news data, which is the above-mentioned video data, is described with reference to FIG. 13. Determine the recognition (S22).

As a result of the above-described determination in step S22, when the still picture is not symmetrical, it is checked whether or not the news icon exists, and if present, the representative image is extracted from the upper right of the quadrant of the screen, that is, the scene is divided using the news icon ( S23) (S24), it reads the next news information and determines whether it is the last (S28). As a result of determination, if it is the last time, it ends, and if it is not the last time, it repeats from the above-mentioned step S21 to generate scene division information.

On the other hand, if the still picture is symmetrical and there is no news icon, the first image of the reporter screen is selected as the representative image, and the next news information is read to determine whether it is the last (S28). As a result of the determination, if it is the last time, the process is terminated. If not, the scene segmentation information is generated by repeating the process from step S21 described above.

Finally, when the news video is to be classified by the subject to generate search information, as shown in FIG. 14, the frequency of the keyword included in the current news caption is calculated (S31).

As a result of calculating the frequency of the keyword in the above-described S31 process, classify the frequency by keyword generating a large number of frequencies (S32), and read the next news information to determine whether it is the last, and if it is the last, the process ends, if not the last. Repeating from the above-described S31.

Second, when the video data input to the scene divider 321 is a movie / video (drama), as shown in FIG. 15, the scene divider 321 may perform a person-by-person search later. According to face recognition, a position at which a face exists is detected (S40).

Then, the position is detected using the actor's voice pattern (S42), the result is synthesized using the above-described face recognition and the voice pattern, the candidate position is detected, the next candidate data is read, and the currently detected candidate position is the last. It is determined whether the position (S44) (S45).

As a result of the above-described determination of step S45, if it is not the last position, the data similar to the candidate data is detected from the database, and the position extraction information for each actor is stored, and then the above-mentioned step S44 is repeated until the last position. To generate search information for a movie / video (drama).

Third, when the video data input to the scene divider 321 is a sport, as shown in FIG. 16, the input video data is divided into video data and audio data (S43-1). Image information is extracted by performing the image processing for each item of the image data divided in the above-described step S43-1 (S43-2), while extracting the word of interest voice pattern from the divided voice data and calculating the voice energy Only data is extracted (S43-3) (S43-4).

The candidate position is extracted by synthesizing the video and audio information extracted through the above-described steps S43-2 and S43-4 (S43-5), and the next candidate data is read to determine the last position of the candidate. (S43-6) (S43-8).

As a result of the above-described determination in step S43-8, when the candidate position is not the last position, the sports highlight search information is generated by extracting a predetermined amount of images before and after the candidate data (S43-9).

In other words, sports analysis is a summary of the entire game. When soccer or baseball is described as an example, soccer produces 2-3 minutes of images before and after the shooting scene occurs. At this time, the extraction method of the point of time when the shot occurs is extracted the position based on the sound signal of the shout of the crowd and the 'shoot' of the announcer and the soccer goal appearing on the screen.

Baseball produces a video of about 2-3 minutes before and after the sound signal and the time when the score occurs, such as the 'hit,' 'home run,' or 'home in' of the announcer. The open subtitles are displayed in consideration of the fact that a corresponding score always appears when a score of a baseball game is generated.

Fourth, when the video data input to the scene divider 321 described above is a show program, as shown in FIG. 17, a position of a face by face recognition is extracted (S51), and a position using a voice pattern for each actor. Extract (S52), extract the pattern section, that is, the starting point and the end point of the song away from the chairman portion (S53).

The position of the candidate is synthesized by synthesizing the position of the face and the voice pattern for each actor, which are extracted through the process S51 and S52 (S54), and the next candidate data is read to determine whether the position of the candidate is the last position (S54). S55) (S56).

As a result of the determination in step S56, when the candidate position is not the last position, it is determined whether the candidate data is stored in the database, and when similar data exists, the similar data is detected and the artist-specific position is extracted and stored. The program search information is generated (S57) (S58).

As described above, when the user executes a communication program installed in the user terminal 200 to the video information search service server 300 which generates search information according to all video contents (S201), first, the video information. As shown in FIG. 18, the search service server 300 transmits a main web page to the user terminal 200 which has established a communication connection, and outputs the main web page on the screen (S202).

If the user inputs user identification information including a user ID and a password through the above-described main web page (S203), and then selects a login button, the video information search service server 300 determines this (S204). The input user identification information is compared with the member information stored in the member DB to determine whether there is matching information.

As a result of the determination, if there is matched information, it is recognized as a registered member (S205), and the desired menu item is selected from the menu item for viewing the video information or the menu item for searching the video information through the aforementioned main web page. (S310 of FIG. 19).

If the user selects a menu item for viewing video information as described above, a web page for outputting corresponding video information on the screen is transmitted to the user terminal 200 (S320 of FIG. 19).

Meanwhile, when the user selects a menu item for searching video information as described above, the web page having the search information as shown in FIG. 21 is transmitted to the user terminal 200 to search for the video group desired by the user. Select it so you can see the scene you want.

In the case of news, the web page having the search information of FIG. 20 includes menu items such as full news information, news information by category, news search, and providing a news chunk video. And menu items for main scene navigator, player search, match rule search, and providing partial video for highlights.In case of movie / video (drama), the main screen is to summarize the drama as the representative screen. There is a menu item that provides drama videos for the actor search box and representative screen.In case of additional information input, there is a menu item for inputting sports additional information and input of additional drama information. , Search by singer, and menu item for each singer's video. In the case of educational broadcasting, a menu item for providing educational broadcasting contents summary, category search, and educational broadcasting contents partial video service is provided.

FIG. 21 is a diagram illustrating a web page of all the news information transmitted to the user terminal when the news video is selected from the main web page transmitted to the user terminal 200. One still image 511 representing the news is shown. -1-5) and representative sentences (511-1-6) representing them, and take the form of enumerating them. At the top, there is a menu to select the type of news (511-1-1), where news is divided into politics, economy, society, culture, international, and others (511-1-4) to easily move to the user's choice. It is possible to use the word you want to search for news (511-1-3) and menu items to view the news by date (511-1-2) provides a menu item to look for only the news you want To help.

The representative screen 511-2-1 of the news corresponding to a certain category and simple information 511-2-2 corresponding thereto are included.

Each of the news representative screens shown on the screen will display a video screen of the selected news by clicking with the mouse.

FIG. 22 is a diagram illustrating a web page transmitted to the user terminal 200 when the user selects the news information menu item for each category from the search information menu item (see FIG. 20) to view news of each category among news video information. There is a part (511-4-3) that shows a video, and above that lists the next news sequentially (511-4-1) so that the user wants to see the news and view the next news. By displaying a representative screen set (511-4-2) and closed caption information (511-4-4) on the news, the news can be understood without having to watch the video.

FIG. 23 is a diagram illustrating a screen transmitted to the user terminal 200 when the user selects the news chunk video information menu item among the menu items of FIG. 20.

Screens for other menu items are also manufactured and operated similarly to the screens described above, and thus detailed description thereof will be omitted.

When the search information generation processor is completed as described above, the user can output only the desired scene on the screen among the video information received by the user terminal and managed by the video information management processor.

That is, when the user wants to see only some scenes of the already broadcasted video information, if the corresponding search information is selected among the search information or the video search information generated by the above-described search information generation processor, the video information output processor may transmit to the video information management processor. Only scenes corresponding to the search information generated by the search information processor are extracted from the corresponding video information managed by the search information processor, and a representative image of each scene is output on the screen, and the user can see only a desired scene.

Accordingly, as described above, the present invention automatically and automatically creates index information by using an automatic analysis system for each program genre such as news, drama, sports, and show, so that a user can easily and conveniently receive a desired video for each scene using the Internet. It is effective to provide.

In addition, the present invention solves the problem that it is difficult to search for a desired part before playing the whole due to the characteristics of the video data, thereby having the effect of finding a desired image by simply selecting a representative representative screen of the user.

In addition, the present invention is different from the method for analyzing the program genre, such as news, drama, sports, shows according to the characteristics of the content in consideration of the user's needs may vary according to the type of content, the news is anchored and It automatically organizes news unit information consisting of reporters to search by articles, dramas build up information by person to enable search by person, and sports automatically builds summary information for important scenes to play long-term matches. The effect is that you can see the highlights without looking.

Although described above with reference to the preferred embodiment of the present invention, those skilled in the art will be variously modified and changed within the scope of the present invention without departing from the spirit and scope of the invention described in the claims below. It will be appreciated.

Claims (21)

Automatically detects and divides scenes of video information to be transmitted through the network according to preset criteria for each field to generate search information on the representative screen, and when a communication connection request is input through the network, the main web page And provide the user with a web page for outputting a video or a web page having video search information in response to a user's request signal input through the main web page so that the user can watch a video or extract a desired scene from the broadcast video. A video information retrieval service server for viewing; And A user performs a communication connection to the video information retrieval service server through a network communication network, outputs a main web page provided by the video information retrieval service server that has made the communication connection on the screen, and inputs the information through the main web page. And a user terminal for receiving a web page for outputting video information or a web page having video search information supplied from a video information search service server according to a request signal, and outputting a representative image of each scene on a screen. Network-based video retrieval system using indexing method. The video indexing service of claim 1, wherein the video information retrieval service server comprises a video input unit configured to receive an image signal, a video analysis unit analyzing the input video data, and a service unit processing the analyzed data. Network-based video retrieval system using the scheme The method of claim 2, wherein the video input unit, A subtitle decoder included in digital broadcast data input from the outside according to the intention of the server operator to extract closed caption information, an MPEG encoder for converting analog broadcast data input into the MPEG data from the outside according to the intention of the server operator, and an external And a controller for checking subtitle information and outputting control signals for caption information detection or MPEG data conversion. The method of claim 2, wherein the video analysis unit, A caption analyzer for adding a time tag generated by using a system clock to the caption information output through the caption decoder, and producing and outputting index material using the time tag-added caption information as a noun unit; and the MPEG encoder A scene divider which automatically generates the divided video information by receiving the output signal and the analog broadcast signal, and generates the additional information according to the index image data made by the caption analyzer and the scene divider and the operation of the server operator. To provide an additional information input device provided to a user, a scene divider for automatically generating video information divided into scenes by MPEG data or digital broadcast data output through the MPEG encoder, and a video broadcasting program through the network communication network. Stream to stream file A network-based video search system using a video indexing system which comprises in one builder. The method of claim 2, wherein the service unit, A service interface for generating a web page having video search information, a web server for transmitting a web page having video search information generated through the service interface to a user through a network communication network, and a stream file producer. A network-based video retrieval system using a video indexing method, characterized in that it comprises a plurality of stream servers for transmitting a stream file to a user through the network communication network. The scene divider of claim 2, wherein the scene divider comprises: A demultiplexing and synchronization module for receiving MPEG bit stream data and dividing it into video information and audio information, and producing a time stamp generated in the table as a table; Using the video image separated by the demultiplexing and synchronization module, the difference value for each frame is calculated between the still images, the difference value is applied to the median filter, and input to the neural network circuit to detect the scene change point. A video segmentation module for separating scenes; Calculate a silent section of the audio data separated by the demultiplexing and synchronization module to determine a silent section as a threshold, and recognize that the start of a new frame when the silent section of the audio data exceeds the threshold. Splitting module; And A video-based video retrieval method using a video indexing method further comprises an information fusion module configured to determine when the scene changes by applying the result of the video splitting module and the result of the audio splitting module to conditional probabilities. system. The scene divider of claim 2, wherein the scene divider comprises: If the type of video information to be input is news, if the closed caption information is present, the position of a predetermined word and the position of the next scene are checked, and if the news video is used, the time based on the anchor's face is checked and the last position is checked. If not, the network-based video retrieval system using the video indexing method characterized in that it generates the news data separated by the news chunk unit by setting the criterion and subtitle synchronization of the image. The scene divider of claim 2, wherein the scene divider comprises: If the type of the input video information is drama, the location of the face by face recognition and the position using the voice pattern for each actor are extracted, the candidate data is extracted by synthesizing the result using the face recognition or the voice pattern, and the candidate data. Extracts and stores the location of each actor by dividing the scene to generate search information by dividing the scene, and the database stores a plurality of face data and voice patterns for each actor, and is based on a network based on a video indexing method. Video search system. The scene divider of claim 2, wherein the scene divider comprises: If the type of video information input is sports, the video and audio are divided, the video information is extracted for each event, the specific words are recognized from the announcer's voice patterns, and the candidate data are extracted using only data with high voice energy. And extracting an image of a predetermined amount before and after the extracted candidate data, generating a highlight scene, and providing the highlight information as search information. The scene divider of claim 2, wherein the scene divider comprises: If the type of video information input is a show program, candidate positions are extracted by using face data, voice patterns for each actor, and pattern sections of the presenter's comments and songs, and the candidate data are detected by synthesizing the face data and the voice patterns of the actors. And checking whether the candidate data exists in the database and segmenting the scene using the recognized artist-specific position to generate search information. The network-based video retrieval system using a video indexing method. (1) The video information retrieval service server checks the field of the video information to be transmitted to the user terminal through a network communication network, and generates the search information by dividing the scene of the video information according to a preset condition corresponding to the checked field. ; (2) the user terminal executing a communication program according to a user's operation to establish a communication connection with the video information retrieval service server; And (3) The video information retrieval service server transmits a main web page for receiving video information or retrieval information to the user terminal according to a user's communication connection request, and the user selects a predetermined number through the main web page received through the user terminal. A method of operating a network-based video retrieval system using a video indexing method, characterized by further comprising the step of receiving a web page having only video information or search information. The method of claim 11, wherein the process of dividing the scene of the moving picture data of the step (1) comprises (1-1) a demultiplexing and synchronization process of dividing MPEG stream data into video stream data and audio stream data; (1-2) a video segmentation step of detecting the change point of the scene by receiving the video stream data divided in the step (1-1), grasping the characteristic of the image on the video screen; (1-3) an audio segmentation process which receives the audio stream data divided in step (1-1), compares it with a preset threshold, detects a silent section, and recognizes it as a transition point of the scene; And (1-4) a data fusion process of detecting the final scene change point using the video stream data divided in the step (1-2) and the audio stream data divided in the step (1-3) and converting the data into data; A method of operating a network-based video retrieval system using a video indexing method characterized in that the data is made. The method of claim 12, wherein the demultiplexing and synchronizing process of the process (1-1), A method of operating a network-based video retrieval system using a video indexing method, characterized by adding time stamp information to each divided video stream data and audio stream data. The method of claim 12, wherein the scene change point is found in the video segmentation process of the process (1-2). (1-2-1) determining whether the input video stream data is the last frame; (1-2-2) extracting only an I picture region of the video stream data when the input video stream data is not the last frame; (1-2-3) extracting a DC image of the extracted I picture; (1-2-4) receiving video stream data having only the extracted DC image and calculating a difference value for each frame between a previous still image and a next still image; (1-2-5) filtering the calculated difference value for each frame through a median filter; (1-2-6) performing a learning process by inputting a difference value for each frame filtered by the median filter to a neural network; And (1-2-7) if the value output through the learning process is greater than or equal to a preset value, registering as a start of a new scene; otherwise, repeating the process (1-2-1). A method of operating a network-based video retrieval system using a video indexing method characterized in that the data is made. The method of claim 11, wherein in the process (1), if the moving picture data is news, (1-i) determining whether closed caption information is input from the outside and extracting a position of a predetermined word when closed caption information is input; (1-ii) extracting a position of a predetermined word and checking a start position of a next scene to determine whether the position is the last position; (1-i) extracting the time when the face of the anchor appears by determining whether news video information is input from the outside; And (1-v) If the result of the determination of step (1-ii) is not the last position or when the time when the anchor face appears in the step (1-v) is extracted, the image information and the caption synchronization are set by the news information. The method of operating a network-based video retrieval system using a video indexing method, characterized by further comprising the step of separating the unit into chunks, and generating search information for searching the separated news chunks. The method of claim 11, wherein the video data is news and the search information is generated by using the representative video extraction method. (1-ⅰ1) calculating a brightness value distribution of the anchor screen; (1-ii1) determining whether the still picture is symmetrical; (1-x1) when the still picture is not symmetrical, extracting a representative image from the upper right of the quadrant of the screen when a news icon exists; And (1-ⅳ1) If the still picture is symmetrical and the news icon does not exist, the video may further include the step of selecting the first image of the reporter screen as a representative image and receiving the next news to generate scene segmentation information. Operation method of network based video retrieval system using indexing method. The method of claim 11, wherein, in the process (1), if the video data is news and is to be classified into subjects to generate search information, (1-ⅰ2) calculating the frequency of the keywords in the current news subtitles; And (1-ii2) A method of operating a network-based video retrieval system using a video indexing method, further comprising: classifying a frequency of a keyword and classifying the keyword by a keyword generating a large frequency. The method of claim 11, wherein, in the process (1), if the moving picture data is a movie / video (drama), (1-A) detecting the location of the face according to face recognition; (1-B) detecting a position using a voice pattern of each actor; (1-C) synthesizing the result using the face recognition and the speech pattern, detecting candidate positions, checking next candidate data, and determining whether the currently detected candidate positions are the last positions; (1-D) As a result of the determination, if it is not the last position, the process similar to the candidate data is detected from the database and the position extraction information for each actor is stored and the process is performed until the final position (1-C). Operating method of a network-based video retrieval system using a video indexing method, characterized in that by repeatedly performing to generate the search information for the drama / movie. The method of claim 11, wherein in the process (1), if the video data is sports, (1-a) dividing the input video data into image data and audio data; (1-b) extracting image information by performing image processing on each item of the divided image data; (1-c) extracting a word of interest speech pattern from the divided speech data, calculating a voice energy, and extracting only high data; (1-d) extracting candidate positions by synthesizing the video and audio information extracted through the steps (1-b) and (1-c); (1-e) reading next candidate data to determine whether the candidate position of step (1-d) is the last position; And (1-f) If the position of the candidate is not the last position, the network-based indexing method further comprises the step of extracting a predetermined amount of images before and after the candidate data and generating sports highlight search information. To operate a video retrieval system. The method of claim 11, wherein if the moving picture data is a show program in the step (1), (1-I) extracting the position of the face by face recognition; (1-Ⅱ) extracting a location using a voice pattern for each actor; (1-Ⅲ) extracting the pattern section of the leader and song; (1-IV) extracting candidate positions by synthesizing the extracted face positions and voice patterns for each actor; (1-V) reading next candidate data to determine whether the candidate position is the last position; And (1-VI) If the position of the candidate is not the last position, it is determined whether the candidate data is stored in the database. If similar data exists, the similar data is detected, and the position of the singer is extracted and stored. A method of operating a network-based video retrieval system using a video indexing method, further comprising generating search information. Checks the field of video information received by the user terminal through a network communication network, generates search information using closed captions according to the checked field, and generates video search information by dividing the scene of the video information. A search information generation processor and a video information management processor which store and manage the information together; And When the user selects the desired search information or video search information from the search information or video search information generated by the search information generation processor, a video indexing method including a video information output processor for extracting and outputting only a corresponding scene from predetermined video information. And a video retrieval program encoded by a computer for reading.