JP2008263481A - Device for extracting video signal section - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a video signal section extracting method and device for extracting video signal section for extracting a particular part of contents desired by an editor in a short period of time, according to the degree of importance. <P>SOLUTION: The video signal section extracting device is provided with a metadata giving part for attaching respective keywords, time base information and importance as metadata to a plurality of sections of a video signal; an in point and out point giving part for setting an in-point and an out-point at a plurality of places of the video signal, and providing a plurality of extraction candidate sections; an importance calculating part for calculating the importance of the extraction candidate sections regarding each of the plurality of extraction candidate sections, on the basis of the importance given to the metadata of each section, corresponding to the extraction candidate sections; and a video signal selecting part for using a result of the importance calculating part, to select a video signal section from the plurality of extraction candidate sections. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a video signal section extracting method and a video signal extracting apparatus for extracting a section of a video signal related to video / audio contents.

  In recent years, in the production of video / audio content (hereinafter also referred to as “content” for short), video segment extraction, for example, for extracting a highlight portion while checking video or audio, is often used.

In general, video segment extraction is performed manually while an editor views content and confirms the content. Therefore, it takes a lot of labor and time to extract the video section.
Therefore, in the case of sports matches, a method has been proposed in which the cheer of the audience is measured and highlight portions are automatically extracted (see Patent Document 1).
JP 2006-180306 A

  However, the method disclosed in Patent Document 1 is not very effective except for sports, and there is also a problem that the start point and end point of the highlight portion are unclear even in sports.

  Accordingly, an object of the present invention is to provide a video signal section extraction method and a video signal section extraction apparatus that solve a conventional problem and extract a specific portion of content desired by an editor in a short time.

In order to achieve the above object, a video signal section extracting device according to the present invention includes a metadata adding section that assigns keywords, time axis information, and importance as metadata to a plurality of sections of a video signal, and the video signal. The in-point / out-point assigning unit that sets in-points / out-points at a plurality of locations and provides a plurality of extraction candidate sections, and for each of the plurality of extraction candidate sections, a meta of each section corresponding to the extraction candidate section An importance calculation unit for calculating the importance of the extraction candidate section based on the importance included in the data, and a video signal selection for selecting a video signal section from a plurality of extraction candidate sections using the result of the importance calculation unit And a section.
With such a configuration, the section desired by the editor can be accurately and quickly extracted from the content.

  According to the video signal section extraction device of the present invention, the section desired by the editor can be accurately and quickly extracted from the content.

  Hereinafter, an embodiment according to the present invention will be described with reference to the drawings.

(Embodiment 1)
FIG. 1 is a block diagram showing a configuration example of a video signal section extraction device according to an embodiment of the present invention. FIG. 2 is a diagram showing an example of scored keyword data that is created by the video signal section extraction device according to the embodiment of the present invention and assigned to a keyword with a score of importance. FIG. 3 is a diagram showing an example of metadata created by the video signal section extraction device according to the embodiment of the present invention, in which scored keyword data and time codes are associated with each other. FIG. 4 shows the sum of the importance levels included in the collective video created by the pointed keyword data, the time code, and the in-point / out-point created by the video signal section extraction device according to the embodiment of the present invention. It is the figure which showed an example of the metadata which linked | related the score.
In FIG. 1, reference numeral 1 denotes a video / audio content playback device such as a camera, VTR, hard disk recorder, semiconductor recorder, optical disk recorder / player, and 2 denotes a video monitor device that monitors a video signal output from the video / audio content playback device 1. Reference numeral 3 denotes an audio monitor device such as a speaker for monitoring the audio signal output from the video / audio content reproduction apparatus 1, and reference numeral 4 denotes time information for inputting time information such as a time code output from the video / audio content reproduction apparatus 1. An input unit 6 is a voice input device such as a microphone for inputting a keyword related to content from the video displayed on the video monitor device 2 and the voice output from the voice monitor device 3, and 7 is from the voice input device 6. A voice recognition unit that recognizes the output voice signal and converts it into keyword data. The keyword dictionary unit 5 stores the keyword data, 5 is an importance level assigning unit for assigning the importance level to the keyword data, 9 is the keyword data output from the speech recognition unit 7, and the importance level output from the importance level giving unit 5 It is a metadata generation part which produces | generates the metadata which linked | related and combined data and the time information data output from the time information input part 4. FIG.
Reference numeral 10 denotes an in-point / out-point adding unit composed of a switch or the like. When operated by the operator 5, the time information input unit 4 gives the respective in-point and out-point giving times.

    Here, the time information input unit 4, the speech recognition unit 7, the keyword dictionary unit 8, the importance level assigning unit 5, and the metadata generation unit 9 are composed of meta data such as voice recognition software, dictionary software, importance level addition software, and metadata generation software. It is built into the computer as software necessary for data creation. As the video / audio content playback device, a playback device such as a television for playing back video / audio signals transmitted from a communication network or a broadcast network can be applied.

  Next, a method for creating metadata will be described.

  The video and audio of the content are viewed on the video monitor device 2 and the audio monitor device 3 to confirm the content, and the audio of the keyword related to the content is input to the audio input device 6. The speech signal of the keyword output from the speech input device 6 is compared with the speech data by the speech recognition unit 7 and the keyword data stored in the keyword dictionary unit 8, and converted to the keyword data closest to the speech data. At this time, the importance level data created by the importance level assigning unit 5 is assigned to the keyword data in advance (FIG. 2), and the corresponding keyword data with importance level data is output from the keyword dictionary 8 to the speech recognition unit 7. Further, the keyword data with importance data output from the voice recognition unit 7 and the time information data output from the time information input unit 4 are input to the metadata generation unit 9. Then, the metadata generation unit 9 associates the time information data with the keyword data with importance data and generates metadata in which both data are combined (FIG. 3).

  As a voice signal to be input to the voice input device 6, any of a voice signal generated by an operator or a voice file format (digital voice signal) created by recording a voice uttered by an operator in a digital recording device in advance is used. obtain.

  As the time information, in addition to the above-described time code, a clock time such as the number of frames or fields of the video signal or the time when the metadata is created may be used.

  As speech recognition methods, for example, Tatsuya Kimura, Akira Ishida, Katsuyuki Niyada: “Unspecified speaker speech recognition for practical use using segment features”, IEICE Journal, J85-D2, 3 pp.398-405 , (2002). Any method that is generally well-known can be applied.

  As the importance level, any number of points, a language indicating the degree (for example, very important, somewhat important, important, etc.), a symbol indicating the degree (for example, AAA, AA, A, B, etc.) can be used. . The importance may be indicated by the number of ☆ marks used for rating hotels and restaurants. Above all, the score is preferable in that a large number of importance levels can be set and the importance level can be intuitively understood.

  Note that some misrecognition may occur in voice recognition. In view of this, an information processing unit (not shown) that corrects audio data when erroneous recognition occurs may be provided in a computer and corrected using a keyboard or a mouse. The metadata generated by the metadata generation unit 9 may be added to the image file according to the purpose, or may be added to a file dedicated to metadata.

  Hereinafter, a method for creating metadata will be described in detail assuming a baseball content as a specific example. In the following explanation, keywords are enclosed in square brackets ("").

  The operator views the content, and utters “player A” when the player A stands at the bat, and utters “runner 3 塁” when the runner is on the heel, for example, when he is on the third heel. Subsequently, the player A utters the content of the attack, for example, the hit direction “right middle”, the hit type “liner”, the attack result “time lease three base”, “1 point”, and the voice of the keyword is input to the voice input device 6. To enter. “Time lease three base” has a high importance level in this content, and therefore, 10 points of importance are given. Therefore, 10 points of importance data are created by the importance level information adding unit 5 and are added to the keyword data of “time lease base” stored in the keyword dictionary unit 8 in advance. Therefore, when the voice recognition unit 7 converts the “time lease three base” voice signal into the keyword data, as shown in FIG. 2, the 10 points of importance data become the “time lease three base” keyword data. Granted automatically. Then, in the metadata generation unit 9, as shown in FIG. 3, the “time lease three base” keyword data to which the 10 points of importance data are added and the “time lease three base” image are added. Metadata associated with the time code (14: 01: 50: 22) is generated.

  By the way, in sports such as baseball, for example, as shown in FIG. 4, the play scene from “player A” entering the batter box to hitting “time lease three base” and putting “1 point” in a lump. Collectively, the total score “16” may be given as the importance data of the entire play scene. Similarly, in the next batch, the total score “14” is given as the importance data of the entire next play scene.

  “[”, “]” Shown in FIG. 4 are symbols indicating an in point (start point) and an out point (end point) of a play scene or the like of a grouped video. In the above description, an input device such as a keyboard, a touch panel, switches, a mouse, a game pad, or the like is used in the above description, but “in”, “in point”, “out”, “out point” are used as voices. Or the like. Depending on the type of sport, “New Play” may be uttered instead of “In” and “In Point”, and “[” may be displayed. Of course, “New Play” may be displayed.

  That is, register “[]”, “In”, “In point”, “]”, “Out”, “Out point” described above to a specific key or switch of the input device, and press the switch to correspond to it. A conversion means for converting to metadata can be provided in the input device, handled in the same way as the voice input means described above, and output to the metadata generation unit 9.

Furthermore, by introducing the rule that the frame immediately before the new “[],“ in ”,“ in point ”is“] ”,“ out ”,“ out point ”,“ [”,“ “]”, “Out”, and “Out Point” can be obtained by inputting only “In” and “In Point”.
In this way, if importance is assigned to each scene grouped together, for example, when creating a news program, it is possible to easily search for important scenes from the content. Will be better.

For example, if the content section selection unit 12 sets the discrimination threshold to 15 points, a section having a total importance of 16 points (upper section in FIG. 4) is selected and output. The section with 14 points of importance (the lower section in FIG. 4) is not selected. There are many non-significant sections (not shown), such as a section of an empty swing and a section of infield fly-out.
For example, if the discrimination threshold is set to 10, not all the sections with the importance 10 or less are selected, but the section with the total importance 16 and the section with the total importance 14 shown in FIG. 4 are selected. As described above, if the discrimination threshold and the importance are designed according to the intention of the editor, a section corresponding to the design value can be arbitrarily selected.

  As described above, as the importance, in addition to the score, a language indicating the degree (for example, very important, slightly important, important, etc.), a symbol indicating the degree (for example, AAA, AA, A, B, etc.), etc. Either can be used. The importance can be indicated by the number of ☆ marks used for rating hotels and restaurants. Furthermore, for each cut in drama production etc., the importance can be shown by OK (indicating the cut that is going to be used), Keep (indicating the cut under consideration), NG (indicating so-called NG cut), etc. . In these cases, somewhat important, AA, ☆☆, OK, etc. are used as threshold values.

    In addition, in the case of video taken for home use, if you want to select a section in which a child is shown as a section to be selected, the In and Out points are provided as described above, and the child is within that section. An application may be considered in which, by calling a name, a section including a child's name is selected as metadata, and a section not calling the child's name is not selected. Furthermore, without adding in / out points, it is possible to simply select one minute before and after the time code when the child's name is called.

    Also, the IN point and OUT point can be used at the recording start point and recording end point.

    That is, the importance level is not limited to the score or the score, but can also be a home run in the case of baseball that can use a speech-recognized keyword itself, or a goal in the case of soccer.

  The video signal section extraction method and extraction apparatus according to the present invention can be used to create metadata related to video / audio contents.

The block diagram which showed the example of 1 structure of the video signal area extraction apparatus of one Embodiment which concerns on this invention The figure which showed an example of the keyword data with importance created with the video signal area extraction device of one embodiment concerning the present invention The figure which showed an example of the metadata produced with the video signal area extraction device of one embodiment which concerns on this invention The figure which showed an example of the other metadata produced with the video signal area extraction device of one embodiment concerning the present invention

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Image | video / audio content reproduction apparatus 2 Image | video monitoring apparatus 3 Audio | voice monitoring apparatus 4 Time information input part 5 Importance assignment part 6 Voice input device 7 Voice recognition part 8 Keyword dictionary part 9 Metadata production | generation part 10 In point / out point provision part 11 Importance calculation part 12 Content section selection part 13 Output part of selected content

Claims (4)

A metadata adding unit that assigns keywords, time axis information, and importance as metadata to a plurality of sections of the video signal,
In points / out points are set in a plurality of locations of the video signal, and a plurality of extraction candidate sections are provided.
For each of the plurality of extraction candidate sections, an importance calculation unit that calculates the importance of the extraction candidate section based on the importance included in the metadata of each section corresponding to the extraction candidate section;
A video signal section extraction device comprising: a video signal selection section that selects a video signal section from a plurality of extraction candidate sections using the result of the importance calculation section. The metadata giving unit
A time information input unit for inputting time information corresponding to the video signal output from the video / audio content playback device;
An audio input device for inputting a keyword corresponding to a video signal by voice;
A keyword dictionary that stores keyword data; and
A voice recognition unit that recognizes a voice signal output from the voice input device and converts it into keyword data stored in the keyword dictionary unit;
A metadata generation unit that generates metadata in which the keyword data output from the speech recognition unit and the time information data output from the time information input unit are associated and combined;
The video signal section according to claim 1, wherein metadata generated by metadata adding means provided with an importance level assigning unit for giving importance level to keyword data stored in the keyword dictionary is used. Extraction device. The video signal section device according to claim 1, wherein the importance is a score. A metadata providing step of assigning keywords, time axis information and importance as metadata to a plurality of sections of the video signal,
In points / out points are set in a plurality of locations of the video signal, and a plurality of extraction candidate sections are provided.
For each of the plurality of extraction candidate sections, an importance calculation step for calculating the importance of the extraction candidate section based on the importance included in the metadata of each section corresponding to the extraction candidate section;
A video signal section extraction method comprising: a video signal selection step of selecting a video signal section from a plurality of extraction candidate sections using the result in the importance calculation step.

Images