JP5056687B2 - Playback apparatus and content playback program - Google Patents

Description

  The present invention relates to a technique for recording and reproducing video content such as a television program. In particular, the present invention relates to a technique for reproducing recorded content in units of scenes.

  When content such as a TV program is reserved for recording, it is generally recorded in units of programs. Therefore, in order to view only a part of the scenes in the program, the user needs to select a program that includes the scene that the user wants to watch, play the program, and skip to the desired scene. As described above, in order to reproduce a predetermined scene in a program, it takes time and effort to view a desired scene.

  Therefore, as a method for extracting a specific scene from a video, for example, a method of extracting a commercial part from a television video, a method of extracting a replay scene, or a method of extracting a part of a weather forecast has been developed. If a scene is extracted using these methods, it is possible to easily reproduce only the scene.

In these methods, a characteristic unique to a scene to be extracted is set, and a desired scene is extracted by searching a part that matches the characteristic from an image.
For example, a technique for storing and storing moving picture information from a key screen for a certain period of time when a key screen is detected from moving picture information based on a received television signal. It is disclosed (for example, Patent Document 1).

In addition, a predetermined moving image display portion is cut out from the input moving image signal to obtain a target moving image signal, and using a feature vector obtained from the target moving image signal, the stored moving image signal recorded in advance is used. A technique for searching for a portion similar to the input moving image signal is also disclosed (for example, Patent Document 2).
Japanese Patent Laid-Open No. 10-42247 JP 2004-193983 A

  According to the above-described conventional technique, since a characteristic unique to a scene to be extracted is set, it is necessary to set a characteristic after a scene to be viewed is determined. It can be used for general needs such as retrieving commercials and weather forecasts. However, in the case where it is not possible to specify in advance what kind of scene should be taken out, such as “I want to take out only the corner that the program is always watching”, the conventional method described above cannot be applied.

  In other words, unless the scene features are known in advance, such as commercials and weather forecasts, it cannot be extracted from the recorded program, and the user can extract and reproduce the desired scene. I can't.

  It is an object of the disclosed playback apparatus to provide a technique capable of extracting a desired scene including a scene whose characteristics are not known in advance.

In order to solve the above-described problem, the disclosed playback device is a playback device that plays back recorded content, and specifies content having a corner composed of scene groups to be played back, and the start position and end position of the corner. A designation unit that receives designation of information, and a first feature amount that divides a designated corner portion of the content designated by the designation unit in units of shots and obtains a first feature amount for each obtained shot An extraction unit; a second feature amount extraction unit that divides another content different from the content in units of shots and obtains a second feature amount for each shot obtained; the first feature amount and the first feature amount; When the first and second feature values fall within a predetermined threshold, a shot having the second feature value is compared with the shot of the corner. To the corner extraction means for extracting as a preparative, and configurations with.

  For a corner designated as a scene group to be extracted from the content, the feature amount (first feature amount) is obtained for each shot. When a new content is input, the content is divided in units of shots, a feature amount (second feature amount) is obtained for each shot, and the feature amount obtained for each shot is compared with the first feature amount. If the value is close to the first feature value, the shot is extracted as a corner. By obtaining a feature amount and extracting a corner, it is possible to extract even a scene group whose features are not known in advance.

  The first and second feature quantity extraction means include a feature group for the image in the corner section, a feature group for the voice in the section, and a time feature group such as the start time and duration of the section. It is good also as extracting the characteristic containing.

  The first feature quantity extraction unit may extract a feature group of a predetermined section before and after the section in addition to the feature group of the section from the start time to the end time of the corner specified by the specifying unit. Good.

The first feature amount extraction unit includes a feature group immediately before the start time of the corner designated by the designation unit, a feature group immediately after the start time, a feature group from the start time to the end time, and a feature group immediately before the end time. Extract the feature group and the feature group immediately after the end time,
For each shot of the other content, the corner extraction means is a probability immediately before the start of the corner, a probability immediately after the start of the corner, a probability of being in the corner, a probability immediately before the end of the corner, and immediately after the end of the corner. It is also possible to calculate a certain probability and extract a section that seems to be the most corner based on the calculation result.

  According to the disclosed playback device, a user can extract a desired scene even if it is a scene whose characteristics are not known in advance.

DESCRIPTION OF EXEMPLARY EMBODIMENTS Hereinafter, preferred embodiments of the invention will be described in detail with reference to the drawings.
<First Embodiment>
FIG. 1 is a diagram illustrating an outline of a corner extraction method according to the present embodiment. In the following, as an example of video content, a TV broadcast program will be described.

  In order to execute the corner extraction method according to the present embodiment, first, it is necessary to record a predetermined program. Here, the program to be subjected to corner extraction is broadcast regularly, and it is necessary to specify in advance so that recording is performed every time by setting a reservation or the like. If there is no recorded program yet, the subsequent processing is executed after the program is recorded for the first time after the reservation.

  As shown in (a), the user designates a corner to be extracted from the recorded programs. In order to allow the user to specify a corner, the recorded video is divided into shot units, and the first frame image of each shot is displayed on the screen in time order. The user designates the frame image of the first shot and the frame image of the last shot through the screen. The first frame to the last frame specified by the user are specified as a corner section. Image and audio features are extracted from the corner sections, and the extracted features are stored in the data storage unit 3 such as a database (DB) as corner features.

  Next, as shown in (b), when a new program is recorded, the newly recorded program is divided into shots, and features are extracted for each shot and stored in the data storage unit 3. Compare with corner features. Based on the comparison, a corner is extracted from the newly recorded program. Information about the corner section extracted from the newly recorded program is stored in the storage unit 3, and the corner feature stored in the storage unit 3, that is, the information stored in (a) is updated as necessary. .

  Then, as shown in (c), when an instruction to reproduce the corner designated by the user is inputted, the corner extracted from the newly recorded program is based on the information stored in the storage unit 3 in (b). Cue the beginning and play the video in the corner section.

  Alternatively, when the corner segment extraction process is executed after the program is recorded, only the corner segment video may be stored, and the original program video may be discarded. In this case, when the user designates reproduction of the video, the video in the corner section is reproduced as it is without performing cueing or the like.

  FIG. 2 is a configuration diagram of the recording / playback apparatus according to the present embodiment. A television program recording / playback apparatus (hereinafter referred to as a recording / playback apparatus) 1 shown in FIG. 2 is connected to a display unit 2 and a data storage unit 3 and includes a scene extraction unit 10. The display unit 2 displays a plurality of shot frames of video content such as a television program, and the user designates the top and end frames of the corner via the display unit 2. The data storage unit 3 stores information related to corner extraction. For example, information about corner characteristics and sections is stored in association with video.

  The scene extraction unit 10 includes a corner section specification unit 11, a corner feature extraction unit 12, a video feature extraction unit 13, a feature matching unit 14, and a corner feature update unit 15, and is specified by the user from content such as a recorded TV program. The extracted corners are extracted from the same TV program recorded thereafter.

  Here, the corner indicates a part of a scene in a certain program. The corner consists of a group of scenes having the same contents in the same program, and is independent of other scenes, and appears every time in TV programs that are broadcast regularly (for example, weekly or daily). However, the contents are different each time.

  Of the scene extraction unit 10, the corner section designation unit 11 starts a corner to be extracted from a program (video content) from which a corner is to be extracted according to information input by the user via the display unit 2. Accepts designation of position / end position. The display unit 2 displays a frame obtained by dividing the video into shots, and the user specifies the positions of the start frame and end frame of the corner to be extracted from the displayed frames. To do. The start and end positions of the corner are specified by, for example, the time and time from the start of the program specified by the user.

  The corner feature extraction unit 12 extracts features about the corner designated by the corner section designation unit 11. Details of the feature extraction method will be described later. The extracted features are stored in the data storage unit 3 as corner features.

  When a video different from the video from which the corner feature is extracted is newly recorded, the video feature extraction unit 13 extracts a feature from the newly recorded video. The feature matching unit 14 compares the corner feature stored in the data storage unit 3 with the video feature extracted by the video feature extraction unit 13, and extracts the corner based on the degree of feature matching. .

  The corner feature updating unit 15 updates information related to the corner feature stored in the data storage unit 3 in accordance with the degree of contribution of the corner feature in the process of extracting the corner from the video in the feature matching unit 14.

  In the recording / playback apparatus 1 shown in FIG. 2, the scene extraction unit 10 extracts features from the video, and if the degree of matching of the features is high compared to the section designated by the user, that is, the corner features, Extract points as corners. The user can view only the corner extracted from the video or archive the corner. Hereinafter, the corner extraction method will be described more specifically.

FIG. 3 is a diagram illustrating an example of features to be extracted. The extracted features can be classified into shot features, corner features, and corner cut features.
As the characteristics of the shot, there are characteristics such as a voice, an image, and a character string as shown in FIG. The voice feature is acquired from the voice waveform of the shot. The feature of the image is acquired from the color and shape of the image, and the character string is acquired from the telop character string included in the shot.

  In the embodiment, in addition to the first shot and the last shot specified by the user, the shot characteristics include the shot immediately before the first shot, the shot after the last shot, and between the first shot and the last shot. Extract from all shots. This is to extract corner breaks, that is, a boundary between a corner to be extracted and another scene with higher accuracy.

  As shown in (b), the corner features include the time from the beginning of the program until the corner starts (start time or start time), the time until the corner ends (end time or end time), and the duration of the corner. Etc.

As a feature of the corner break, for example, as shown in (c), there is a feature that a commercial flows before and after the corner.
When the first shot and the last shot are designated by the user, a corner break is determined based on the designated first shot and last shot. The feature of the corner to be extracted is obtained by obtaining the time from the beginning of the program to the first shot and the time from the last shot to the end of the corner, respectively. Then, based on the obtained corner start time and end time, the feature about the necessary shot is acquired.

  As for the features of shots, the features such as the above-mentioned voice, image, and character string are represented by multidimensional vectors. Note that a technique for expressing shot features with a multidimensional vector uses a known technique, and thus a detailed description thereof will be omitted. First, among the recorded programs, the vector is obtained for a shot or the like designated by the user, and “corner characteristics” are obtained. As described above, in the embodiment, not only the first shot and the last shot specified by the user but also the shot immediately before the first shot, the next shot after the last shot, All shots from the first shot to the last shot.

  When a new program is recorded, the video of the program is divided into shots, a feature vector is calculated for each shot, and the difference from the corner feature is calculated, for each shot of the newly recorded program The distance from the previously obtained feature vector is calculated as the “corner feature”. The distance between the corner feature and the feature vector for the shot of the program from which the corner is to be extracted is defined as “corner shot likelihood” for the shot. In the following description, among the “corner features”, for example, the shot likelihood of the corner obtained from the difference from the feature vector acquired from the first shot is “first shot likelihood”.

Next, a calculation method of “corner shot likelihood” will be described with reference to FIG.
FIG. 4 is a diagram for explaining a method of calculating the shot likelihood of a corner. Here, a case where the likelihood of the first shot is calculated will be described as an example.

  As described above, the corner feature 22 for the first shot acquired from the recorded program and the feature 21 for a shot obtained by dividing the program are both expressed as multidimensional vectors. When the feature 21 and the corner feature 22 for a shot are similar, the vector difference (distance) is small, and when the feature 21 and the feature 22 are not similar, the vector difference is large. .

  In the embodiment, the difference between these feature vectors is weighted and added for all items, and a value obtained by dividing the total by the number of items is set as the head shot likelihood. As for the weight, “1.0” is set for all items in the expected state as shown in FIG. For items that have contributed greatly to corner extraction, the weight is increased, and for items that have a small contribution, the weight is decreased. The weight update process will be described later.

  When the “corner shot likelihood” is obtained for each shot, it is determined whether or not each shot of the newly recorded program belongs to the corner section using the obtained corner shot likelihood.

  FIG. 5 is a diagram illustrating a method for calculating the likelihood of a corner section. (A) in the figure shows the likelihood of a shot related to the corner section for each shot of a certain program. Here, the shot identification numbers are given in time series, and in the example shown in the figure, the program includes seven shots (shot 1 to shot 7). In addition to the shot quality, the elapsed time counted from the start time of the shot program is also described. For example, the shot 1 is from the start time of the program to 15 seconds (0: 00-0: 15).

  In this embodiment, two different shots are selected from a plurality of shots constituting a program, and whether or not the two selected shots are the first shot and the last shot of the corner is obtained by the above method. Judgment is made using “shot-likeness”.

The corner section likelihood is calculated by the following formula. In the following calculation formula, it is assumed that shot m and shot n (m <n) are the first shot and the last shot, respectively.
(Like a corner section of shots m to n)
= (Like-shot-likeness of shot m-1)
+ (Looks like the first shot of shot m)
+ (Sum of shots m + 1 to n-1)
+ (Like shot last shot)
+ (Like shot after final shot n + 1)
+ (Difference between start time of corner feature and start time of shot m)
+ (Difference between corner feature end time and shot n end time)

For example, regarding the likelihood of the corner section of shot 1 to shot 2 in (b), since shot 0 immediately before shot 1 does not exist, the value “500” set in the embodiment as “unreliable value”. Is entered. Since no shot exists between shot 1 and shot 2, the value of the term of intermediate shot is “0”.

  The start time of the corner feature is 30 seconds after the start of the program as shown in (a), whereas the start time of shot 1 is the same as the start of the program (after 0 seconds), and the difference is 30 Seconds. As a result, the value “30” is input to the “start time difference” in the calculation formula for the corner section likelihood. The end time of the corner feature is 2 minutes or 120 seconds after the start of the program, whereas the end time of shot 2 is 25 seconds after the start of the program. As a result, “120−25 = 95” is input for the end time difference.

  Similarly, a combination of the first and last shots is obtained for all shots of the program, and the likelihood of a corner section is calculated from the above formula for each combination. A shot group in a corner section is extracted with a combination of shots having the smallest value of the corner section as the first and last shots. In the example shown in FIG. 5, since the “likeness of corner section” of shots 2 to 5 is the smallest, shots 2 to 5 are extracted as corner sections.

When a shot group is extracted as a corner section, the weight of the corner feature is updated according to the degree of contribution to the specification of the corner section.
FIG. 6 is a diagram illustrating a method of updating the corner feature weights. The right side of (a) of the figure shows a table for corner features stored in the data storage unit 3 of FIG. The corner feature table stores a corner feature and a weight associated with each feature.

  First, it is determined whether or not the value of “corner section likelihood” is equal to or less than a predetermined threshold. In the example of FIG. 5, the corner section likelihood of shots 2 to 5 is determined. If the corner section likelihood is equal to or less than a predetermined threshold, the corner feature is updated.

  As a method of updating the feature value, for example, an average of the corner feature value stored in the table of the data storage unit 3 and the shot feature value in the corner section is set as a new corner feature value. In addition to this, a method may be considered in which the number of shot features that have been used for updating the corner features so far is stored, and the average is calculated by weighting the corner features accordingly.

For the weight, for example, there is a method of updating according to the following formula (calculation formula 23 in FIG. 6).
(New weight) = (old weight) + (old weight) × (threshold−distance) / (threshold × learning coefficient)
Here, the threshold value is “200” in the embodiment. The learning coefficient in the above equation is a value that determines how much the weight is updated. The larger the learning coefficient, the slower the update. In the example illustrated in FIG. 6, the learning coefficient is “5”.

  It is considered that the smaller the distance, the more contributed to the determination of the corner section. For this reason, the weight update calculation formula 23 is set such that the feature having a small distance has a large weight when the next corner section is determined, and the feature having a large distance has a small weight. .

  According to the above method, since the corner feature is updated to the average of the extracted shot feature, the extracted shot feature and the updated corner feature are more mutually compared than before the update. The value is approaching. For this reason, if the distance (feature vector) is obtained after the corner feature is updated for shots having the same feature, the distance is smaller than that before the update.

  Further, when the shot-likeness (first shot-likeness in the example of FIG. 6) is calculated by reflecting the updated weight, it is as shown in (b), and the shot-likeness value is smaller than that before the update. I understand that. In other words, after the corner feature (and its weight) is updated, the shot-likeness can be calculated with higher accuracy, and thereby extraction of the corner section can be performed with higher accuracy.

FIG. 7 is a flowchart showing processing for extracting a corner section.
First, in step S1, shots divided for a newly recorded video are acquired in order from the top of the video. At this time, the extracted scene features are also acquired. In step S2, it is determined whether or not a shot has been acquired. If it is determined that it has not been acquired, it is determined that the processing for all shots has been completed, and the process proceeds to step S7.

  If it is determined in step S2 that the shot has been acquired, the process proceeds to step S3, and the feature for the first shot among the corner features is acquired from the data storage unit 3 in FIG. Then, in step S4, the feature of the shot acquired in step S1 is compared with the feature of the first shot taken out from the data storage unit 3 in step S3, the likelihood of the first shot is calculated, and the process proceeds to step S5.

  In step S <b> 5, the feature for the last shot among the corner features is acquired from the data storage unit 3. In step S6, the feature of the shot acquired in step S1 is compared with the feature of the final shot extracted from the data storage unit 3 in step S5, and the likelihood of the final shot is calculated.

After that, although not shown in FIG. 7, the likelihood of the front shot, the likelihood of the last shot, and the intermediate shot likelihood of the corner are also calculated in the same manner as described above.
If all the shots of the newly recorded video have been calculated for the first shot, the last shot, the first front shot, the last last shot, and the intermediate shot, the process proceeds to step S7. Then, using the “shot-likeness” obtained in the above step, the corner-likeness is calculated, a group of shots in the corner-section is extracted, and the process ends. The method for calculating the likelihood of the corner section and extracting the shot group belonging to the corner section is as described above.

FIG. 8 is a flowchart showing processing for updating corner features.
First, in step S11, shots determined to be corner sections are acquired in order. As described above, here, not only the shots in the corner section from the first shot to the last shot, but also the shots used for calculating the likelihood of the corner section (pre-start shot and post-end shot) are acquired.

In step S12, it is determined whether or not a desired shot has been acquired. If it is determined that the desired shot has not been acquired, no particular process is performed, and the process ends.
If it is determined in step S12 that a desired shot has been acquired, the process proceeds to step S13, and the likelihood of a corner section for the acquired shot is acquired. Here, for example, the value of the “shot-first shot likelihood” value is a value for a shot that is determined to be the previous shot at the beginning of the corner (in the example of FIG. 5, the shot-first shot likelihood “5” in the example of FIG. 5) For the value of “first shot likelihood”, the value of the corner interval calculated using the value of the shot determined to be the first shot of the corner (in the example of FIG. 5, the likelihood of the first shot of shot 2 is “10”) is acquired. To do.

In step 14, it is determined whether or not the acquired value of the corner section likelihood is equal to or less than a predetermined threshold value. If the value of the corner section likelihood exceeds the threshold value, it is determined that the shot has not contributed to the determination of the corner section, the corner feature and the weight are not updated, and the process returns to step S11.

  In step S14, when the value of the shot likelihood of the corner is equal to or smaller than the threshold value, the process proceeds to step S15, and the distance between the shot feature and the corresponding corner feature is acquired in order. In step S16, it is determined whether or not distances have been acquired for all features. If there are features for which distances have not been acquired, the process returns to step S11.

  If it is determined in step S16 that distances have been acquired for all the features, the process proceeds to step S17 to update the corner features and weights. The update method is as described above with reference to FIG.

  As described above, according to the recording / playback apparatus 1 according to the present embodiment, a corner to be extracted is designated in advance, and the feature of the corner is acquired and held as a corner feature. The corner feature is acquired for each shot related to a period designated as a corner by the user. Then, when taking out the corner from the newly recorded video, each shot of the newly recorded video is compared with each of the owned corner features, and the parts with similar features in the video are compared at the same corner. Extract as there is. When an instruction to reproduce the corner designated by the user is input, the video of the extracted corner section is reproduced from the newly recorded video. Even in a scene in which the feature (scene) of the content video is not known in advance, it is possible to extract a section close to the feature specified by the user as a corner section.

<Second Embodiment>
The recording / playback apparatus 1 according to the present embodiment is the same as the above embodiment in that the video for which the corner section is to be extracted (newly recorded video) is divided in units of shots. The above is performed in that the head of the corner section is searched in order, and if the head is found, it is determined whether the next and subsequent shots are shots in the corner section or the last shot. Different from form.

  Since the configuration and the like of the recording / playback apparatus according to this embodiment are the same as those of the above-described embodiment, description thereof will be omitted, and different points will be mainly described.

FIG. 9 is a flowchart showing a corner section extraction method according to the second embodiment.
First, in step S21, shots are designated in order from the beginning of the program video, and the characteristics of the designated shot are acquired. In step S22, it is determined whether or not the shot feature has been acquired. If it is determined that the shot feature has not been acquired, it is determined that there is no corner-like section, and the process ends.

  If it is determined in step S22 that the shot feature has been acquired, the process proceeds to step S23, and the feature for the first shot is acquired from the corner features stored in the data storage unit 3 in FIG. In step S24, the feature of the shot acquired in step S21 is compared with the feature of the first shot extracted from the data storage unit 3, and the likelihood of the first shot is calculated. In step S25, it is determined whether or not the obtained first shot likelihood is equal to or less than a predetermined threshold value. If it is determined that the likelihood of the first shot exceeds a predetermined threshold, the process returns to step S1 and the search for the first shot is continued.

  If it is determined in step S25 that the likelihood of the first shot is equal to or less than a predetermined threshold value, the process proceeds to step S26, and a feature is acquired for the shot next to the shot acquired in step S21.

  In step S27, it is determined whether or not the shot feature has been acquired. If it is determined that the shot feature could not be acquired, it is determined that the corner section has not been found, and the process ends.

  If it is determined in step S27 that the shot feature has been acquired, the process proceeds to step S28, and the feature for the final shot is acquired from the data storage unit 3. In step S29, the feature of the shot acquired in step S26 is compared with the feature of the final shot extracted from the data storage unit 3, and the likelihood of the final shot is calculated. In step S30, it is determined whether the obtained final shot likelihood is equal to or less than a predetermined threshold value.

  If it is determined in step S30 that the final shot likelihood is equal to or less than the predetermined threshold, it is determined that the shot acquired in step S26 is the final shot, and the process proceeds to step S34.

  If it is determined in step S30 that the final shot likelihood exceeds a predetermined threshold, the process proceeds to step S31, and the intermediate shot likelihood for the shot acquired in step S26 is further determined.

  Specifically, in step S31, the characteristics of the intermediate shot are acquired from the data storage unit 3. In step S32, the intermediate shot likelihood of the shot acquired in step S26 is calculated. In step S33, it is determined whether or not the obtained intermediate shot likelihood is equal to or less than a predetermined threshold value.

  In step S33, if the likelihood of an intermediate shot exceeds the threshold value, it is determined that the shot is not a corner section shot, and the process returns to step S21. In step S21, a feature is acquired for the shot next to the shot in which the feature is acquired in step S26, and the above processing is repeated.

  If it is determined in step S33 that the likelihood of an intermediate shot is less than or equal to the threshold value, it is determined that the shot is an intermediate shot in a corner section, and the process returns to step S26. In step S26, a feature is acquired for the next shot after the previously acquired feature, and the above process is repeated.

  As described above, if it is determined in step S30 that the final shot likelihood is equal to or less than the predetermined threshold value, the process proceeds to step S34, and the shot that is acquired in step S21 and determined that the first shot likelihood exceeds the threshold value is determined in the corner section. The corner section is obtained using the first shot, the shot acquired in step S26 and determined that the likelihood of the final shot exceeds the threshold value as the final shot, and the process ends.

  In the above method, the corner section is extracted by comparing the likelihood of the first shot, the likelihood of the final shot, and the likelihood of the intermediate shot with a predetermined threshold. However, the method of extracting the corner section is not limited to this. For example, a configuration may be adopted in which the corner section is obtained by using the above-mentioned front pre-shot and final post-shot.

  In addition, as the threshold value to be used, the same value may be set for each of the head shot-likeness, the last shot-likeness, and the intermediate shot-likeness, or different values may be set.

As described above, as in the corner section extraction method according to the present embodiment, it is determined whether or not it is the first shot in order from the top of the video, and when the first shot is obtained, the next and subsequent shots are determined. The corner section can also be extracted by a method of determining whether the shot is the intermediate shot or the final shot. Similar to the above-described embodiment, when an instruction to reproduce the corner designated by the user is input, the video of the extracted corner section is reproduced from the newly recorded video.

  Incidentally, the recording / playback apparatus 1 of FIG. 2 can be configured using, for example, an information processing apparatus (computer) as shown in FIG. 10 includes a CPU (central processing unit) 1001, a memory 1002, an input device 1003, an output device 1004, an external storage device 1005, a medium drive device 1006, and a network connection device 1007, which are connected to each other via a bus 1008. It is connected.

  The memory 1002 includes, for example, a read only memory (ROM), a random access memory (RAM), and the like, and stores programs and data used for processing. The CPU 1001 performs necessary processing by executing a program using the memory 1002.

  The data storage unit 3 in FIG. 2 corresponds to the memory 1002. 2 is realized by executing a program stored in the memory 1002. The corner section specifying unit 11, the corner feature extracting unit 12, the video feature extracting unit 13, the feature matching unit 14, and the corner feature updating unit 15 shown in FIG. Corresponds to the function.

  The input device 1003 is, for example, a keyboard, a pointing device, a touch panel, etc., and is used for inputting instructions and information such as designation of a corner section from the user of the recording / playback device 1. The output device 1004 is, for example, a display or the like, and is used for outputting alarms, processing results, and the like.

  The external storage device 1005 is, for example, a magnetic disk device, an optical disk device, a magneto-optical disk device, a tape device, or the like. The information processing apparatus stores the program and data in the external storage device 1005 and loads them into the memory 1002 for use as necessary.

  The medium driving device 1006 drives a portable recording medium 1009 and accesses the recorded contents. The portable recording medium 1009 is an arbitrary computer-readable recording medium such as a memory card, a flexible disk, a CD-ROM (compact disk read only memory), an optical disk, and a magneto-optical disk. The operator stores the program and data in the portable recording medium 1009 and loads them into the memory 1002 for use as necessary.

  The network connection device 1007 is connected to an arbitrary communication network such as a local area network (LAN) or the Internet, and performs data conversion accompanying communication. The information processing apparatus receives the program and data from an external apparatus via the network connection apparatus 1007 as necessary, and loads them into the memory 1002 for use.

  FIG. 11 shows a computer-readable recording medium that can supply a program and data to the information processing apparatus of FIG. Programs and data stored in the portable recording medium 1009 and the database 1103 of the server 1101 are loaded into the memory 1002 of the information processing apparatus 1102. The server 1101 generates a carrier signal for carrying the program and data, and transmits the carrier signal to the information processing apparatus 1102 via an arbitrary transmission medium on the network. The CPU 1001 executes the program using the data and performs necessary processing.

Regarding the above embodiment, the following additional notes are disclosed.
(Appendix 1)
A playback device for playing back recorded content,
A specifying means for receiving a content having a corner made up of a group of scenes to be reproduced, and specifying information specifying a start position and an end position of the corner;
A first feature amount extraction unit that divides a designated corner portion of the content designated by the designation unit in units of shots and obtains a first feature amount for each obtained shot;
A second feature amount extraction unit that divides another content different from the content in units of shots and obtains a second feature amount for each obtained shot;
When the first feature value and the second feature value are compared, and the first and second feature values fall within a predetermined threshold, a shot having the second feature value is selected as the corner. Corner extraction means for extracting as a shot of,
A playback apparatus comprising:
(Appendix 2)
The first and second feature quantity extraction means include a feature group for the image in the corner section, a feature group for the voice in the section, and a time feature group such as the start time and duration of the section. The reproduction apparatus according to appendix 1, characterized in that a feature including: is extracted.
(Appendix 3)
The first feature quantity extraction unit extracts a feature group of a predetermined section before and after the section in addition to a feature group of a section from the start time to the end time of the corner specified by the specifying unit. The reproducing apparatus according to appendix 2.
(Appendix 4)
The first feature amount extraction unit includes a feature group immediately before the start time of the corner designated by the designation unit, a feature group immediately after the start time, a feature group from the start time to the end time, and a feature group immediately before the end time. Extract the feature group and the feature group immediately after the end time,
For each shot of the other content, the corner extraction means is a probability immediately before the start of the corner, a probability immediately after the start of the corner, a probability of being in the corner, a probability immediately before the end of the corner, and immediately after the end of the corner. 4. The playback device according to appendix 3, wherein the most likely corner section is extracted based on a certain probability.
(Appendix 5)
The first feature amount extraction unit extracts a feature group immediately after the start time of the corner designated by the designation unit and a feature group immediately before the end time, respectively.
The corner extraction means searches for a shot immediately after the start of the corner based on the probability that it is immediately after the start of the corner from the beginning of the shot of the other content, and for shots after the shot immediately after the start of the corner, 4. The playback apparatus according to appendix 3, wherein a shot of a section that seems to be the most corner is extracted by searching for a shot immediately before the end of the corner based on a probability immediately before the end of the corner.
(Appendix 6)
As a result of comparing the first and second feature amounts in the corner extraction unit, a corner feature update unit that increases the weight of the first feature amount for feature amounts that are close to each other;
The reproduction apparatus according to appendix 1, further comprising:
(Appendix 7)
A content playback program for playing back recorded content,
Accepts content with a corner consisting of scenes to be played and specification of information specifying the start and end positions of the corner,
Of the specified content, the specified corner portion is divided into shot units,
Obtain the first feature value for each obtained shot,
Dividing other content different from the content in units of shots, obtaining a second feature amount for each obtained shot,
When the first feature value and the second feature value are compared, and the first and second feature values fall within a predetermined threshold, a shot having the second feature value is selected as the corner. A content reproduction program that causes a computer to execute a process of extracting as a shot of

It is a figure explaining the outline | summary of the extraction method of a corner. It is a block diagram of the recording / reproducing apparatus which concerns on embodiment. It is a figure which shows the example of the characteristic to extract. It is a figure explaining the calculation method of the shot quality of a corner. It is a figure explaining the method of calculating corner section likeness. It is a figure explaining the method to update the weight of a corner feature. It is the flowchart (the 1) which showed the process which extracts a corner area. It is the flowchart which showed the process which updates a corner feature. It is the flowchart (the 2) which showed the process which extracts a corner area. It is a block diagram of information processing apparatus. It is a figure which shows a recording medium.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Recording / reproducing apparatus 2 Display part 3 Data storage part 10 Scene extraction part 11 Corner area designation | designated part 12 Corner feature extraction part 13 Image | video feature extraction part 14 Feature matching part 15 Corner feature update part

Claims (3)

A playback device for playing back recorded content,
A specifying means for receiving a content having a corner made up of a group of scenes to be reproduced, and specifying information specifying a start position and an end position of the corner;
A first feature amount extraction unit that divides a designated corner portion of the content designated by the designation unit in units of shots and obtains a first feature amount for each obtained shot;
A second feature amount extraction unit that divides another content different from the content in units of shots and obtains a second feature amount for each obtained shot;
When the first feature value and the second feature value are compared, and the first and second feature values fall within a predetermined threshold, a shot having the second feature value is selected as the corner. and a corner extracting means for extracting a shot,
The first feature quantity extraction unit includes a feature group in a section immediately before the start time of the corner designated by the designation unit, a feature group in a section immediately after the start time, and a feature group in a section from the start time to the end time , To extract the feature group of the section immediately before the end time and the feature group of the section immediately after the end time,
For each shot of the other content, the corner extraction means is a probability immediately before the start of the corner, a probability immediately after the start of the corner, a probability of being in the corner, a probability immediately before the end of the corner, and immediately after the end of the corner. A reproduction apparatus characterized by calculating a certain probability and extracting a section most likely to be a corner based on the calculation result . The first and second feature quantity extraction means include a feature group for the image in the corner section, a feature group for the voice in the section, and a time feature group such as the start time and duration of the section. The reproduction apparatus according to claim 1, wherein a feature including the feature is extracted. A content playback program for playing back recorded content,
The content reproduction program is
Accepts content with a corner consisting of scenes to be played and specification of information specifying the start and end positions of the corner,
Of the specified content, the specified corner portion is divided in units of shots, and a first feature amount is obtained for each obtained shot.
Dividing other content different from the content in units of shots, obtaining a second feature amount for each obtained shot,
When the first feature value and the second feature value are compared, and the first and second feature values fall within a predetermined threshold, a shot having the second feature value is selected as the corner. Let the computer execute the process of extracting as a shot of
The processing for obtaining the first feature value includes a feature group of a section immediately before the start time of a designated corner, a feature group of a section immediately after the start time, a feature group of a section from the start time to the end time, and an end time Extract the feature group of the section immediately before and the feature group of the section immediately after the end time,
The process of extracting as a shot of the corner includes, for each shot of the other content, the probability immediately before the start of the corner, the probability immediately after the start of the corner, the probability of being in the corner, the probability immediately before the end of the corner, and the corner A content reproduction program characterized by calculating a probability of being immediately after the end of the video and extracting a section that seems to be the most corner based on the calculation result .