JP5513197B2 - Content evaluation apparatus, method, program, and integrated circuit - Google Patents

Description

  The present invention relates to a technology for evaluating content such as still images, moving images, and music.

  In recent years, content that can be viewed by users has been increasing against the background of digitization of broadcasting and multi-channel broadcasting. In addition, with the increase in capacity of recording media such as hard disks, the amount of content that can be held by users is also increasing.

Therefore, regardless of whether or not to watch, it has become possible to view all of the broadcast content once and then only view the content that you want to watch later.
However, there is a problem that it takes time and effort to select a desired content from a huge amount of recorded contents. Therefore, it is necessary to evaluate each content and recommend a highly evaluated content so that the user can efficiently search for the content that the user wants to see.

  As a conventional technique for evaluating and recommending content, when viewing content, a list of genres is displayed to allow the user to select the genre of the content, and the user's preference for each genre is determined based on information on the selected genre. Patent Document 1 discloses a technique for evaluating and recommending other content based on the determined preference level.

  Also, as a technique that can be used to evaluate and recommend content, in photo printing, the user selects a favorite image, detects preference data from the selection result by the user, and based on the detected preference data A technique for selecting a photograph to be printed out is disclosed in Patent Document 2.

  Also, Japanese Patent Application Laid-Open No. 2004-151867 discloses a technique for creating a profile indicating a viewing tendency of a program for each user in a broadcast receiving system and recommending a program that matches the created profile.

JP 2006-229678 A JP 2007-072823 A JP 2009-239686 A

  However, in any of the above-described methods, it is necessary to perform selection operations such as selecting a content genre, selecting a favorite image, selecting a program to be viewed, etc. prior to content evaluation. Take the trouble.

  An object of the present invention is to provide an apparatus for performing evaluation for recommending content without allowing the user to perform a selection operation in advance without the selection operation.

  In order to solve the above-described problem, a content evaluation apparatus according to the present invention includes a sample content acquisition unit that acquires sample content that is content used for evaluation of content to be evaluated, and all or part of the sample content acquired by the sample content acquisition unit. Attempts to detect each of a plurality of features from the sample content, and a feature detection amount specifying means for specifying a detection amount that is the amount of the sample content in which the feature is detected for each feature, and each feature corresponding to one content to be evaluated And an evaluated content evaluation means for evaluating the evaluated content based on the detection amount specified by the feature detection amount specifying means.

  Here, the evaluated content (Evaluate Content) is content to be evaluated. The content to be evaluated is, for example, video and audio of a broadcast program, a still image or video downloaded from the Internet, a photograph received from a third party, or the like.

  The sample content is content used for evaluating the evaluated content, and is content different from the evaluated content. The sample content is, for example, a photograph or a movie owned by the user.

  The feature indicates what kind of content the evaluated content and the sample content include. For example, “includes an image corresponding to an object such as an automobile”, “high brightness of the image”, “includes sound of a specific pattern”, and the like can be considered.

  Each feature has detection reference information for detecting the feature from the evaluated content and the sample content, and the respective detection criteria are different from each other.

  The content evaluation apparatus having the above configuration is characterized by the sample content file acquired by the user's input if the content to be evaluated is a broadcast program and the sample content is a photo or video file captured and stored by the user The detected amount of each feature is specified based on the detected result, the feature is detected from the broadcast program, and the broadcast program is evaluated based on the detected amount of the feature detected from the broadcast program.

  Since it is considered that the user captures a lot of objects of high interest, the detection amount of each feature is considered to reflect the user's interest. Therefore, the content evaluation apparatus can evaluate the broadcast program according to the user's interest in each feature.

  By using this content evaluation apparatus and selecting a broadcast program having a high evaluation from among many broadcast programs, the user can watch a broadcast program with high interest to the user.

  At this time, the user does not need to bother selecting the necessary files from the stored files, and only has to input all the files that are held together. As a result, it takes less time to find a broadcast program that suits the user's preference.

  In addition, since the feature used for evaluation is detected from the data of the photos and videos stored in the file without using additional information such as metadata at least for the sample content file, the user can There is no need to add metadata such as information and tags. This further reduces the time and effort required to find a broadcast program that meets the user's preference.

Here, the feature detection amount specifying means may calculate the feature detection amount based on a sum of values indicating the degree to which each sample content in which the feature is detected includes the feature.
By doing in this way, it is possible to obtain the detection amount by placing emphasis on the feature that appears strongly in the sample content, not just the number detected. Therefore, an evaluation result that better reflects the user's interest can be obtained.

Since a user's high interest is large and is likely to be taken for a long time, a recommendation result that better reflects the user's interest can be obtained.
Here, further, the sample content is a still image, each feature represents each object included in the still image, and the feature detection amount specifying means includes one sample content in which one feature is detected, A value indicating the degree to which the feature is included may be calculated as a ratio of the area of the portion in which the object represented by the feature is included in the sample content to the area of the sample content.

By doing so, it becomes possible to evaluate a broadcast program with emphasis on the feature that the user shows in the photograph.
Here, the sample content is a moving image, and each feature represents each object included in a frame in the moving image, and the feature detection amount specifying means detects each feature using a specific type of frame as a detection target. For one sample content in which one feature is detected, for each frame including the object represented by the feature of the sample content, the area of the portion where the object is included in the frame with respect to the area of the frame The ratio is calculated, and the value indicating the degree to which the sample content includes the feature is calculated as the ratio of the sum of the ratios for each frame including the object of the sample content to the number of frames that are the detection target of the sample content. It may be calculated as

By doing in this way, it becomes possible to evaluate a broadcast program with emphasis on the feature that the user has greatly captured in the moving image.
Here, the sample content is a moving image, and each feature represents each object included in a frame in the moving image, and the feature detection amount specifying means detects each feature using a specific type of frame as a detection target. Attempting to detect the value of the sample content in which one feature is detected, the value indicating the degree to which the sample content includes the feature, the number of frames in which the object represented by the feature of the sample content is included, and the detection target of the sample content It may be calculated as a ratio to the number of frames.

In this way, broadcast programs can be evaluated with emphasis on the features that the user has taken in the video for a long time.
Here, the evaluated content evaluation means may evaluate the evaluated content based on the sum of the detected amount values of the features corresponding to one evaluated content.

By doing in this way, it becomes possible to highly evaluate a broadcast program including many features that are of high interest to the user.
Here, the sample content is content to which information on the time of creation, acquisition, or update is given, and the feature detection amount specifying unit evaluates the given time by the evaluated content evaluation unit. It may be possible to try to detect features from sample content that falls within a range from time to a predetermined period.

If the predetermined period is 7 days, the broadcast program can be evaluated based on the features that the user is interested in in the last 7 days.
Here, the sample content is content to which information on the time of creation, acquisition, or update is given, and the feature detection amount specifying unit evaluates the given time by the evaluated content evaluation unit. From the sample content that enters the first period that is a range from the time to a predetermined period before, and the sample content that enters the second period in which the assigned time is a predetermined range in the past from the first period Attempts to detect features, and the evaluated content evaluation means has a detection amount specified by the sample content whose time falls within the first period, and each time corresponding to one evaluated content has a time of the second period The content to be evaluated may be evaluated based on a change with respect to the detection amount specified by the sample content entering.

  Assume that the first period is one day on the day of evaluation and the second period is one day seven days before the evaluation, and the broadcast program is evaluated based on the feature that the user's interest has increased in the last seven days. Will be able to.

  Here, the evaluated content is content to which metadata is added, and the content evaluation device further includes means for associating the metadata of the evaluated content with the characteristics, The evaluated content may be evaluated based on the detected amount of each feature corresponding to the metadata of the evaluated content.

  Assuming that the metadata is keyword text indicating program content, the feature can be detected by simple text comparison, so that the broadcast program can be evaluated at high speed.

  In addition, since it is considered that appropriate keywords are added by broadcasters who know the contents of the program, evaluation that reflects the contents of the program more appropriately than when mechanical features are detected mechanically by image processing Results are obtained.

It is a figure which shows the usage example of this invention. It is a figure which shows the function structure of the program recommendation apparatus which concerns on Embodiment 1. FIG. It is a figure which shows the data structure and example of a content of a sample content group. It is a figure which shows the data structure and content example of feature detection reference | standard information. It is a figure which shows the data structure and content example of characteristic importance information. It is a figure which shows the data structure and content example of a to-be-evaluated content group. It is a figure which shows the data structure and content example of to-be-evaluated content characteristic information. It is a figure which shows the data structure and content example of to-be-evaluated content evaluation information. It is a figure which shows the example of the program which the program recommendation apparatus selected. It is a figure which shows the example which a display apparatus displays the program which a program recommendation apparatus recommends. It is a flowchart which shows the process at the time of acquiring sample content. It is a figure which shows the image which detects a feature from an image. It is a flowchart which shows the process which updates the detection amount of a feature. It is a flowchart which shows the process which calculates the importance of a feature. It is a flowchart which shows the process which evaluates a to-be-evaluated content group. It is a flowchart which shows the process at the time of detecting a characteristic from to-be-evaluated content. It is a flowchart which shows the process which evaluates one to-be-evaluated content. It is a flowchart which shows the process which a program recommendation apparatus recommends a program. It is a figure which shows the example of the content of the characteristic importance information in Embodiment 2. It is a figure which shows the data structure and content example of the characteristic detection reference | standard information of Embodiment 2. FIG. It is a figure which shows the example of the evaluation result of the to-be-evaluated content in Embodiment 2. FIG. It is a figure which shows the example of the program selected in Embodiment 2. FIG. It is a figure which shows the image which specifies the detection area of the characteristic with a still image and a moving image. 10 is a flowchart illustrating processing for updating a feature detection amount according to the second exemplary embodiment. 10 is a flowchart illustrating processing for evaluating evaluated content in the second exemplary embodiment. It is a figure which shows the image of the time change of the detection amount of a feature. It is a figure which shows the data structure of the sample content group in Embodiment 3, and an example of content. 14 is a flowchart illustrating processing when acquiring sample content in the third embodiment. 14 is a flowchart illustrating processing when a feature is detected from content to be evaluated in the third embodiment. 14 is a flowchart illustrating a process in which a program recommendation device recommends a program in the third embodiment. 12 is a flowchart illustrating processing for specifying importance levels of all features in the third embodiment. It is a figure which shows the image of the time change of the detection amount of a feature. It is a figure which shows the data structure and content example of the characteristic importance information in Embodiment 4. 10 is a flowchart illustrating processing for specifying importance levels of all features in the fourth embodiment. 10 is a flowchart illustrating processing for updating a feature detection amount in the fourth embodiment. 10 is a flowchart illustrating processing for calculating feature importance in the fourth embodiment. It is a figure which shows the data structure and content example of the to-be-evaluated content group in Embodiment 5. FIG. It is a figure which shows the data structure of the feature detection reference | standard information in Embodiment 5, and the example of content.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
<1. Embodiment 1>
Embodiment 1 of the content evaluation apparatus according to the present invention is a program recommendation apparatus that evaluates evaluated content based on features detected by image processing from evaluated content and sample content, and recommends highly evaluated evaluated content. Will be explained.

Here, the content to be evaluated is a moving image content of a program broadcast on a wired or wireless basis.
The sample content is assumed to be files of still images and moving image data that are personally stored and stored in a recording medium such as a memory card.

The feature indicates what kind of contents the evaluated content and the sample content include, and a plurality of features may be detected from one evaluated content or sample content. Here, it is assumed that each feature indicates the content that an object such as a car or a flower is reflected in a still image and a moving image.
<1-1. Overview>
FIG. 1 is an image diagram of a usage example of the program recommendation device 100 according to the first embodiment of the present invention, including its peripheral devices.

  In the example of the figure, the program recommendation apparatus 100 acquires a plurality of evaluated contents from the broadcast 101 and a plurality of sample contents from the memory card 104, and displays a highly evaluated evaluated content among the acquired evaluated contents. It outputs to 102 and recommends. This recommendation operation is triggered by a user instruction using the controller 105.

  The display device 102 is a device that displays the video output from the program recommendation device 100. Here, the display device 102 is a digital television that displays video by being connected to the program recommendation device 100 via a cable such as a High-Definition Multimedia Interface (HDMI) cable.

  The memory card 104 is a recording medium that records sample content. The program recommendation device 100 can attach and detach the memory card 104, and can acquire the sample content stored in the memory card 104.

  The controller 105 is a device for operating the program recommendation device 100, and transmits an operation signal to the program recommendation device 100 by communication using infrared light, for example. The user can use the controller 105 to select a program to be viewed from among programs recommended by the program recommendation device 100.

  When the user inserts the memory card 104 into the program recommendation device 100 and inputs the sample content, the program recommendation device 100 acquires the sample content from the memory card 104, and each feature that the program recommendation device 100 can detect from each acquired sample content Try to detect. Then, the detection amount and importance of the feature are specified by the number of sample content files from which each feature is detected, and the specified detection amount and importance are recorded.

  When the program recommendation device 100 acquires the evaluated content from the broadcast 101, the program recommendation device 100 tries to detect each feature from the acquired evaluated content. Then, the content to be evaluated is evaluated based on the sum of recorded importance levels of the features detected from the content to be evaluated, and the evaluation value of the evaluation result is recorded.

When the user transmits a signal indicating a recommendation instruction to the program recommendation device 100 by the controller 105, the program recommendation device 100 that has received the signal descends from the highest evaluation value according to the evaluation value of each content to be evaluated. Three evaluated contents are selected. Then, the selected three contents to be evaluated are output and displayed on the display device 102.
<1-2. Configuration>
Hereinafter, the configuration of the program recommendation device 100 will be described by dividing it into a hardware configuration and a functional configuration.
<1-2-1. Hardware configuration>
The program recommendation apparatus 100 according to the first embodiment includes, as a hardware configuration, a receiving antenna and input terminal for acquiring content to be evaluated, a memory card reader for acquiring sample content, an HDMI output terminal for outputting video, and a controller. A receiver for receiving the optical signal, a memory for storing data and a program, and a processor for executing the program.
<1-2-2. Functional configuration>
FIG. 2 is a block diagram showing a functional configuration of the program recommendation device 100.

  In the figure, the program recommendation device 100 includes a sample content acquisition unit 201, a sample content storage unit 202, a feature detection reference storage unit 203, a feature detection amount specifying unit 204, an evaluated content acquisition unit 208, an evaluated content storage unit 209, The evaluation content evaluation unit 210, the evaluated content evaluation value storage unit 214, the operation input unit 215, the content selection unit 216, and the output unit 217 are configured.

Further, the feature detection amount specifying unit 204 includes a sample content feature detection unit 205, a feature importance calculation unit 206, and a feature importance storage unit 207.
The evaluated content evaluation unit 210 includes an evaluated content feature detection unit 211, an evaluated content feature storage unit 212, and an evaluated content evaluation value calculation unit 213.

  Sample content acquisition unit 201, sample content feature detection unit 205, feature importance calculation unit 206, evaluated content acquisition unit 208, evaluated content feature detection unit 211, evaluated content evaluation value calculation unit 213, operation input unit 215, content The operations of the selection unit 216 and the output unit 217 are realized by the processor executing a program stored in the memory.

The sample content storage unit 202, the feature detection reference storage unit 203, the feature importance storage unit 207, the evaluated content storage unit 209, the evaluated content feature storage unit 212, and the evaluated content evaluation value storage unit 214 are realized by a memory. .
<1-2-3. Explanation of each functional component>
Below, each function structure part of Embodiment 1 shown in FIG. 2 is demonstrated.

  The sample content acquisition unit 201 has a function of acquiring a sample content file recorded in the memory card 104. Further, a sample content ID 301 corresponding to each sample content is generated, and a set of the acquired sample content and the sample content ID 301 corresponding to each sample content is stored in the sample content group 300 stored in the sample content storage unit 202. It also has a function to add.

The sample content storage unit 202 is a memory area for storing the sample content group 300 acquired by the sample content acquisition unit 201.
FIG. 3 shows a data configuration and example contents of the sample content group 300 stored in the sample content storage unit 202. As shown in the figure, the sample content group 300 includes a sample content ID 301 and a sample content 302.

  In the example of the figure, the sample content ID 301 from “SC0001” to “SC0004” in the sample content group 300 and the image of the sample content 302 having the sample content ID 301 are shown. Although not shown in the example, the sample content 302 of the sample content group 300 includes a moving image.

The feature detection reference storage unit 203 is a memory area for storing feature detection reference information 400 having a plurality of feature detection references 402, which are references for detecting features from still images and moving images.
FIG. 4 shows a data configuration and example contents of the feature detection reference information 400 stored in the feature detection reference storage unit 203. As shown in the figure, the feature detection reference information 400 includes a feature ID 401 and a feature detection reference 402.

  In the example of the figure, among the feature detection reference information 400, a feature ID 401 from “FE0001” to “FE0004” and a feature detection reference 402 of the feature having the feature ID 401 are shown. In the example, the feature detection reference 402 is written in descriptive terms such as “automobile features”. However, the feature detection reference 402 is actually composed of data such as feature value values used to detect the features.

  The feature detection amount specifying unit 204 tries to detect a feature from the sample content 302 stored in the sample content storage unit 202 using the feature detection reference 402 stored in the feature detection reference storage unit 203, and from the result of the detection attempt. It has a function of specifying the detection amount 501 of each feature. Also, the importance 502 is calculated from the detected amount 501 of each feature.

  The sample content feature detection unit 205 included in the feature detection amount specifying unit 204 uses a feature detection reference 402 stored in the feature detection reference storage unit 203 from the sample content 302 stored in the sample content storage unit 202. Has a function to try detection.

  The feature importance calculation unit 206 included in the feature detection amount specifying unit 204 includes a sample content 302 in which the feature is detected for each feature based on the result of the sample content feature detection unit 205 attempting to detect the feature from the sample content 302. The number of detected files 501 is obtained as the number of files. Furthermore, it has a function of obtaining the importance 502 of each feature from the detected amount 501 of each feature. Each feature has a function of storing a set of feature ID 401, detection amount 501, and importance 502 as feature importance information 500 in the feature importance storage unit 207.

Here, the feature detection amount 501 is obtained as the number of sample content files in which the feature is detected.
The feature importance 502 is obtained as a value obtained by dividing the feature detection amount 501 by the maximum value of all feature detection amounts 501.

The feature importance storage unit 207 included in the feature detection amount specifying unit 204 is a memory area for storing the feature importance information 500 obtained by the sample content feature detection unit 205.
FIG. 5 shows a data configuration and example contents of the feature importance information 500 stored in the feature importance storage unit 207. As shown in the figure, the feature importance level information 500 includes a feature ID 401, a detection amount 501, and an importance level 502.

  In the example of the figure, among the feature importance information 500, the feature ID 401 from “FE0001” to “FE0004”, the feature detection amount 501 and the importance 502 having the feature ID 401 are shown.

  The evaluated content acquisition unit 208 acquires a broadcast program as evaluated content using a receiving antenna or an input terminal, generates an evaluated content ID 601 corresponding to each evaluated content, and the acquired evaluated content and its content It has a function of adding a set with an evaluated content ID 601 corresponding to each evaluated content to the evaluated content group 600 stored in the evaluated content storage unit 209.

The evaluated content acquisition unit 208 acquires a broadcast program as needed and stores it in the evaluated content storage unit 209.
The evaluated content storage unit 209 is a memory area for storing the evaluated content group 600 acquired by the evaluated content acquisition unit 208.

  FIG. 6 shows the data structure and example contents of the evaluated content group 600 stored in the evaluated content storage unit 209. As shown in the figure, it is composed of an evaluated content ID 601 and an evaluated content 602.

  In the example of the figure, the evaluated content ID 601 from “EC0001” to “EC0004” and the evaluated content 602 having the evaluated content ID 601 in the evaluated content group 600 are shown. Note that the content to be evaluated 602 is actually a moving image, but since it cannot be expressed on paper, it is shown as a picture of one scene representing each moving image.

  The evaluated content evaluation unit 210 attempts to detect a feature from the evaluated content 602 stored in the evaluated content storage unit 209 using the feature detection reference 402 stored in the feature detection reference storage unit 203, and then tries to detect the feature. The evaluated content 602 is evaluated based on the result and the importance 502 of the feature specified by the feature detection amount specifying unit 204, and the evaluation value 801 indicating the evaluation of the evaluated content 602 corresponds to the evaluated content 602 evaluated. The set with the evaluated content ID 601 is stored in the evaluated content evaluation value storage unit 214.

  The evaluated content feature detection unit 211 included in the evaluated content evaluation unit 210 includes the feature detection reference 402 stored in the feature detection reference storage unit 203 from the evaluated content 602 stored in the evaluation content storage unit 209. Has a function of trying to detect a feature. In addition, a set of the evaluated content ID 601 corresponding to the evaluated content 602 attempted to be detected and the feature IDs 401 of all the features actually detected from the evaluated content 602 is evaluated in the evaluated content feature storage unit 212. It also has a function of storing it as content feature information 700.

  The evaluated content feature storage unit 212 included in the evaluated content evaluation unit 210 is a memory area for storing the evaluated content feature information 700 generated by the evaluated content feature detection unit 211.

  FIG. 7 shows the data structure and example contents of the evaluated content feature information 700 stored in the evaluated content feature storage unit 212. As shown in the figure, the evaluated content feature information 700 includes an evaluated content ID 601 and a feature ID 401.

  In the example of the figure, among the evaluated content feature information 700, some of the features detected from the evaluated content ID 601 from “EC0001” to “EC0004” and the evaluated content 602 having the evaluated content ID 601. ID 401 is shown.

  The evaluated content evaluation value calculation unit 213 included in the evaluated content evaluation unit 210 includes the evaluated content feature information 700 stored in the evaluated content feature storage unit 212 and the features specified by the feature detection amount specifying unit 204. It has a function of calculating an evaluation value 801 indicating an evaluation for the evaluated content from the importance 502 calculated based on the detection amount 501. In addition, it has a function of storing a set of the evaluated content ID 601 of the evaluated content for which the evaluation value 801 is calculated and the calculated evaluation value 801 as the evaluated content evaluation information 800 in the evaluated content evaluation value storage unit 214.

Here, the evaluation value 801 of the evaluated content is calculated as the sum of the importance 502 of each feature detected from the evaluated content.
The evaluated content evaluation value storage unit 214 is a memory area for storing the evaluated content evaluation information 800 generated by the evaluated content evaluation unit 210.

  FIG. 8 shows a data configuration and example contents of the evaluated content evaluation information 800 stored in the evaluated content evaluation value storage unit 214. As shown in the figure, the evaluated content evaluation information 800 includes an evaluated content ID 601 and an evaluation value 801.

  In the example of the figure, among the evaluated content evaluation information 800, an evaluated content ID 601 from “EC0001” to “EC0004” and an evaluation value 801 that is an evaluation of the evaluated content having the evaluated content ID 601 are shown. Yes.

  The operation input unit 215 receives an optical signal corresponding to an operation such as selecting content by the content selection unit 216 from the controller 105 using a light receiver or the like, and sends a command or data corresponding to the received signal to the content selection unit 216. It has a function.

  In response to the command received from the operation input unit 215, the content selection unit 216 recommends and outputs three content items to be evaluated and recommended in descending order from the highest evaluation value 801 stored in the evaluated content evaluation value storage unit 214. A function of selecting the evaluation content ID 601 is provided. The selected content to be evaluated is output by a content selection unit 216 described later.

FIG. 9 shows an example of evaluated content selected as content recommended by the content selection unit 216.
The example in the figure shows that an evaluated content having an evaluated content ID 601 of “EC0001”, “EC0065”, and “EC0002” is selected. For the selection of the evaluated content, it is sufficient to select only the evaluated content ID 601. However, in this example, the picture of the evaluated content 602 and the evaluation value 801 are also shown in order to facilitate the image.

  The output unit 217 has a function of causing the display device 102 to display a screen for recommending the evaluated content 602 corresponding to the evaluated content ID 601 selected by the content selection unit 216 on an output interface such as an HDMI output terminal.

FIG. 10 is an example of an image output from the output unit 217.
In the example shown in the figure, a reduced image of one evaluated content 602 out of three recommended evaluated content 602, program information corresponding to the evaluated content 602, and a value of an evaluation value 801 are displayed. Yes.
<1-3. Data>
Next, data handled by the program recommendation device 100 having the above-described configuration will be described in detail.
<1-3-1. Sample content group>
The sample content group 300 is a set of a set of a sample content 302 used for evaluating the evaluated content and a sample content ID 301 corresponding to each sample content 302.

The sample content group 300 is generated from the sample content file acquired by the sample content acquisition unit 201 and stored in the sample content storage unit 202.
The sample content ID 301 is an identifier (IDentifier) for uniquely identifying each sample content within the program recommendation device 100. Here, it is assumed that the sample content ID 301 is assigned a four-digit serial number from 0001 in the order in which the sample content acquisition unit 201 acquires the sample content 302, and “SC” is added to the head of the number. For example, the sample content ID 301 of the 100th acquired sample content 302 is “SC0100”.

In the following description, the sample content shown in FIG. 3 will be described as a specific example, and the sample content 302 and the data corresponding to the sample content 302 will be referred to as a sample content ID 301. For example, the sample content 302 identified by the sample content ID 301 of “SC0001” is referred to as “SC0001 sample content 302”.
<1-3-2. Feature detection criteria information>
The feature detection reference information 400 is data that is stored in the feature detection reference storage unit 203 and used for feature detection by the sample content feature detection unit 205 and the evaluated content feature detection unit 211. Here, it is assumed that the feature detection reference information 400 is prepared in advance and stored in the feature detection reference storage unit 203, and is not generated, changed, or deleted during the operation of the program recommendation device 100.

  The component feature ID 401 of the feature detection reference information 400 is an identifier for uniquely identifying each feature within the program recommendation device 100. Here, for the feature ID 401, it is assumed that a four-digit serial number is assigned sequentially from 0001 in the order in which the feature detection reference storage unit 203 stores the feature detection reference information 400, and "FE" is added to the head of the number. . For example, the feature ID 401 of the 100th feature is “FE0100”.

  In the following description, when the feature shown in FIG. 4 is described as a specific example, the feature and the data corresponding to the feature one-to-one are referred to as a feature ID 401. For example, the feature identified by the feature ID 401 of “FE0001” is referred to as “feature of FE0001”, and the feature detection reference 402 serving as a reference for detecting the feature of “FE0001” is referred to as “feature detection reference 402 of FE0001”.

  The feature detection reference 402 of the component of the feature detection reference information 400 is data used by the sample content feature detection unit 205 and the evaluated content feature detection unit 211 to detect a feature, and detection corresponding to each feature. It is the data which shows the reference | standard of.

The feature detection reference 402 is configured by data indicating representative values and ranges such as SIFT (Scale-Invariant Feature Transform) feature amounts and color information indicating objects corresponding to the respective features.
<1-3-3. Feature Importance Information>
The feature importance information 500 is information indicating how much each feature has been detected from the sample content group 300, and is used for evaluation of the evaluated content performed by the evaluated content evaluation value calculation unit 213. The feature importance level information 500 is generated by the feature importance level calculation unit 206, stored in the feature importance level storage unit 207, and used by the evaluated content evaluation unit 210.

  The feature detection amount 501 of the component of the feature importance information 500 is the amount of sample content for which the sample content feature detection unit 205 has detected the feature. Here, it is assumed that the number of files of the sample content 302 in the sample content group 300 in which the feature can be detected.

  The feature importance 502 of the constituent elements of the feature importance information 500 is a value obtained by normalizing the detection amount 501 so that the maximum value becomes 1 while maintaining the ratio of the detection amounts 501 between the features. . That is, the importance 502 of a feature is a value obtained by dividing the feature detection amount 501 by the maximum detection amount 501 of all feature detection amounts 501.

When the sample content included in the sample content group 300 increases, the detection amount 501 also increases as a whole. Therefore, even if the detection amount 501 is large, it cannot be distinguished whether the detection amount 501 is large because the user's interest is high or the detection amount 501 is large because the sample content is large. Here, when the importance level 502 is used, all the importance levels 502 fall within the range of 0 to 1 regardless of the total amount of the sample content.
<1-3-4. Content to be evaluated>
The evaluated content group 600 is a set of a set of evaluated content 602 to be evaluated and evaluated content ID 601 corresponding to each sample content 302.

The evaluated content group 600 is generated from the evaluated content acquired by the evaluated content acquisition unit 208 and stored in the evaluated content storage unit 209.
The evaluated content ID 601 that is a component of the evaluated content group 600 is an identifier for uniquely identifying each evaluated content within the program recommendation device 100. Here, it is assumed that the evaluated content ID 601 is assigned a 4-digit serial number from 0001 in the order in which the evaluated content acquisition unit 208 acquires the evaluated content 602, and “EC” is added to the head of the number. For example, the evaluated content ID 601 of the evaluated content 602 acquired 100th is “EC0100”.

In the following description, when the content to be evaluated shown in FIG. 6 is described as a specific example, the content to be evaluated and the data corresponding to the content to be evaluated are referred to as an evaluated content ID 601. For example, the evaluated content 602 identified by the evaluated content ID 601 of “EC0001” is called “EC0001 evaluated content 602”.
<1-3-5. Evaluated content feature information>
The evaluated content feature information 700 is data indicating features corresponding to each evaluated content.

  The evaluated content feature information 700 is generated by the evaluated content feature detection unit 211, stored in the evaluated content feature storage unit 212, and used by the evaluated content evaluation value calculation unit 213.

A plurality of features may be detected from the evaluated content 602, or none may be detected. When one or more are detected, all the detected features are stored in the evaluated content feature information 700. When none is detected, data indicating that none is detected is evaluated content feature information. 700 is stored.
<1-3-6. Evaluated content evaluation information>
The evaluated content evaluation information 800 is information indicating evaluation of the evaluated content. The evaluated content evaluation information 800 is generated by the evaluated content evaluation value calculation unit 213, stored in the evaluated content evaluation value storage unit 214, and used by the content selection unit 216.

The evaluation value 801 constituting the evaluated content evaluation information 800 is evaluated by the evaluated content evaluation value calculation unit 213 for each evaluated content based on the evaluated content feature information 700 and the feature importance information 500. .
<1-4. Operation>
Next, the operation of the program recommendation device 100 according to the present invention will be described.

The operation of the program recommendation device 100 is roughly divided into three processes: a process for detecting features from the evaluated content, a process for acquiring sample content, and a process for recommending a program.
The processing for acquiring the sample content is executed when the user inputs the sample content by inserting the memory card 104 into the program recommendation device 100 or the like. In this process, the sample content stored in the memory card 104 is acquired, the feature is detected from each sample content 302, and the feature importance information 500 is specified using the result.

  Processing for detecting features from the evaluated content is periodically executed. In this process, a feature is detected and recorded from each evaluated content stored in the evaluated content storage unit 209, and each evaluated content is evaluated from the detection result.

  The process of recommending a program is executed when the user performs an operation for viewing a recommended program using the controller 105 and the program recommendation device 100 receives the operation signal. In this process, a predetermined number (e.g., three) of the evaluated contents 602 is selected in descending order from the one having the highest evaluation value 801, and recommendation is performed by displaying the selected contents.

Hereinafter, each of the three processes will be described with reference to the flowchart of the drawing.
<1-4-1. Processing when obtaining sample content>
In the processing for acquiring the sample content, the sample content that can be acquired is acquired, and detection of the feature from each acquired sample content 302 is attempted. Then, the importance 502 of each feature is specified based on the result, and each content to be evaluated that has already been recorded is evaluated using the importance 502 of each specified feature.

In the following, a process when the program recommendation device 100 acquires sample content will be described using the flowchart of FIG.
First, the sample content acquisition unit 201 determines whether there is unacquired sample content that can be newly acquired (step S1101).

  If there is one (step S1101: Yes), the sample content acquisition unit 201 acquires one of the unacquired sample contents 302. Then, a sample content ID 301 is generated, and a set of the sample content ID 301 and the acquired sample content 302 is added to the sample content group 300 stored in the sample content storage unit 202 (step S1102).

Hereinafter, the sample content 302 acquired here will be referred to as sample content x.
Next, the sample content feature detection unit 205 attempts to detect each feature from the sample content x using the feature detection reference 402 stored in the feature detection reference storage unit 203 (step S1103).

Then, the feature importance calculation unit 206 performs an update process of the feature detection amount 501 detected from the sample content x (step S1104).
When the update processing of the feature detection amount 501 ends, the process returns to step S1101 in order to acquire the next sample content 302.

  If it is determined in step S1101 that there is no unacquired sample content that can be newly acquired (S1101: No), the feature importance level calculation unit 206 calculates the importance level 502 of each feature (step S1101). S1105).

Finally, the evaluated content evaluation value calculation unit 213 performs evaluation processing of the evaluated content group based on the calculated importance 502 of each feature (step S1106).
<1-4-1-1. Feature detection>
Hereinafter, processing in which the sample content feature detection unit 205 attempts to detect features from the sample content 302 will be described.

  When the sample content 302 to be processed is a still image, the feature can be detected using the conventional technique. For example, as shown in FIG. 12, a local feature amount is extracted from a rectangular area called a detection window for one still image 1200 and matched with a feature amount pattern for each feature (objects such as flowers and automobiles). There is a method called object detection for detecting the presence of the object.

  For example, if there are a pattern that matches a flower image and a pattern that matches a car image as a reference for detecting features, a rectangular area 1201a in the still image 1200 can be From the rectangular area 1201b, a feature that an automobile is present is detected.

  If the sample content 302 to be processed is a moving image, all the frames included in the moving image are regarded as still images, and feature detection is attempted. If it is detected from at least one of all the frames, the feature is detected.

Taking the case where the still image 1200 of FIG. 12 is one frame in the sample content x as an example, the feature of the flower and the feature of the car are detected from the sample content x.
Note that, when attempting to detect a feature from the evaluated content 602, the feature detection is attempted by the same method.
<1-4-1-2. Update processing of feature detection amount>
The update process of the feature detection amount 501 is a process for updating the detection amounts 501 of all the features detected from one sample content 302.

This processing is performed by the feature importance calculation unit 206, and the feature importance information 500 as a result thereof is stored in the feature importance storage unit 207.
In the following, a process of updating the feature detection amount 501 with the features detected by the feature importance calculation unit 206 from one sample content will be described using the flowchart of FIG. Here, the sample content for which the feature is detected is referred to as sample content x.

  First, of the features detected from the sample content x, it is determined whether there is a feature whose detection amount 501 has not yet been updated since the update processing of the feature detection amount 501 is started (step S1301). If there is, the process proceeds to the following process (step S1301: Yes). If not, the update processing of the feature detection amount 501 is terminated (step S1301: No).

  One of the features whose detection amount 501 has not yet been updated is selected, and the value of the detection amount 501 of the selected feature is updated to a value that is one greater than the original detection amount 501 (step S1302).

The updated feature detection amount 501 is stored in the feature importance storage unit 207 (step S1303).
When step S1303 ends, the process returns to step S1301.
<1-4-1-3. Feature importance calculation processing>
The feature importance 502 calculation process is a process for calculating the importance 502 of all the features.

This processing is performed by the feature importance calculation unit 206, and the feature importance information 500 as a result thereof is stored in the feature importance storage unit 207.
Hereinafter, a process in which the feature importance level calculation unit 206 calculates the importance level 502 of each feature will be described using the flowchart of FIG.

  First, it is determined whether or not there is a feature whose importance 502 has not yet been calculated since the feature importance 502 calculation processing is started among the features whose detection amount 501 has been updated (step S1401). If there is, the process proceeds to the following process (step S1401: Yes). If not, the calculation process of the feature importance 502 is terminated (step S1401: No).

  One of the features whose importance level 502 has not yet been calculated is selected, and the value of the importance level 502 of the selected feature is divided by the detection amount 501 of the feature by the largest detection amount 501 among the detection amounts 501 of all the features. The calculated value is calculated (step S1402).

The calculated feature importance 502 is stored in the feature importance storage unit 207 (step S1403).
When step S1403 ends, the process returns to step S1401.
<1-4-1-4. Evaluation processing of evaluated content group>
The evaluation processing for the evaluated content group 600 is processing for evaluating all the evaluated content 602.

This process is performed by the evaluated content evaluation value calculation unit 213, and the evaluated content evaluation information 800 as a result is stored in the evaluated content evaluation value storage unit 214.
Hereinafter, the evaluation processing of the evaluated content group 600 performed by the evaluated content evaluation value calculation unit 213 will be described using the flowchart of FIG.

First, although not shown in the figure, all the evaluated contents are set to an unevaluated state. This can be realized, for example, by setting the evaluation value 801 to an invalid value.
Then, it is determined whether there is unevaluated content to be evaluated (step S1501). If yes (step S1501: Yes), the process proceeds to the following process. If not (step S1501: No), the evaluation process for the evaluated content group is terminated.

One of the unevaluated contents to be evaluated is selected, and the selected contents to be evaluated are evaluated (step S1502). This evaluation process will be described in detail later.
When the evaluation of the selected content to be evaluated ends, the process returns to step S1501.
<1-4-2. Processing when detecting features from evaluated content>
The process for detecting a feature from the evaluated content attempts to detect the feature from each of the evaluated content 602 stored in the evaluated content storage unit 209. Then, the evaluated content is evaluated by using the importance 502 specified by the feature detection amount specifying unit 204 for each detected feature.

Processing when the program recommendation device 100 detects a feature from the evaluated content will be described with reference to the flowchart of FIG.
First, it is determined whether there is an evaluated content that has not yet been subjected to feature detection (step S1601). If there is any (step S1601: Yes), the following processing is performed. If not (No at step S1601), the process ends.

  In the following description, processing is performed with one of the evaluated contents that have not yet been subjected to feature detection as a feature detection target, and the evaluated content 602 that is the detection target is referred to as evaluated content y.

  First, the evaluated content feature detection unit 211 detects a feature from the evaluated content y (step S1602), and sets a set of the evaluated content ID 601 of the evaluated content y and the detected feature feature ID 401 to the evaluated content feature storage unit. 212 (step S1603).

Subsequently, an evaluation process for the content to be evaluated y is performed (step S1604).
When the evaluation process for the evaluated content y is completed, the process returns to step S1601 in order to detect features from the next evaluated content.
<1-4-2-1. Evaluated content evaluation process>
The evaluated content evaluation process is a process of evaluating the evaluation value 801 of the evaluated content 602 as the sum of the importance 502 of each feature detected from one evaluated content 602.

  The evaluated content evaluation processing is performed by the evaluated content evaluation unit 210, and the evaluated content evaluation information 800 as a result is stored in the evaluated content evaluation value storage unit 214.

  Hereinafter, a process when the evaluated content evaluation unit 210 evaluates one evaluated content will be described using the flowchart of FIG. Here, one evaluated content to be evaluated is referred to as evaluated content y.

First, the evaluation value of the evaluated content y is set to 0 (step S1701).
Subsequently, it is determined whether there is a feature that has not been used for calculation of the evaluation value 801 after the evaluation process of the evaluated content y is started among the features detected from the evaluated content y (step S1702). ). If there is (step S1702: Yes), the following processing is performed.

  One of the features not yet used for calculating the evaluation value 801 of the evaluated content y is selected, and the importance 502 of the selected feature is added to the evaluation value 801 of the evaluated content y (step S1703).

When the process of step S1703 ends, the process returns to step S1702.
If there is no feature that has not yet been used to calculate the evaluation value 801 of the evaluated content y (step S1702: No), the set of the evaluated content ID of the evaluated content y and the evaluation value 801 of the evaluated content y Is stored in the evaluated content evaluation value storage unit 214 as the evaluated content evaluation information 800 (step S1704).

  When the evaluated content evaluation information 800 having the same evaluated content ID 601 already exists in the evaluated content evaluation value storage unit 214, the old evaluated content evaluation information 800 is overwritten with the new evaluated content evaluation information 800.

When step S1704 ends, the evaluation process for the evaluated content y ends, and the process returns to the caller process.
<1-4-3. Processing to recommend programs>
The process of recommending a program is a process of recommending by selecting and displaying three contents to be evaluated having a high evaluation value 801 in response to user input.

This processing is performed by the content selection unit 216, and the result is displayed by the output unit 217.
Hereinafter, the process of recommending a program performed by the content selection unit 216 will be described with reference to the flowchart of FIG.

First, it waits until there exists operation which requests | requires the recommendation by a user (step S1801: No). If there is an operation requesting recommendation, the following processing is performed (step S1801: Yes).
Using the evaluated content evaluation information 800 stored in the evaluated content evaluation value storage unit 214, three evaluated contents are selected in descending order from the one having the highest evaluation value 801 (step S1802).

Next, the selected content to be evaluated is displayed on the output unit 217 and recommended (step S1803).
When the recommendation process is completed, the process returns to step S1801 and waits again until there is a next request.
<2. Second Embodiment>
Next, as the second embodiment, the feature can be detected in the program recommendation device 100 according to the first embodiment, where the evaluation value and the detection amount are obtained depending on whether the feature is detected from the evaluated content and the sample content. A modified program recommendation apparatus in which the operation is changed so as to obtain the evaluation value and the detection amount by weighting using the area of the part will be described.

In the following, the description will focus on the part corresponding to the difference from the program recommendation device 100 according to the first embodiment.
<2-1. Overview>
The modified program recommendation device evaluates the evaluation value 801 by weighting and adding the importance 502 of each feature detected from the evaluated content by weighting the area of the detected portion of the feature. .

  Further, the modified program recommendation device counts the amount of sample content in which the feature is detected by weighting the importance of the feature 502 by the area of the portion in which the feature is detected from each sample content.

Here, the area is the number of pixels included in the whole or a part of the image.
<2-2. Configuration>
The hardware configuration of the modified program recommendation device is exactly the same as the hardware configuration of the program recommendation device according to the first embodiment.

The functional configuration of the modified program recommendation device is also substantially the same as the functional configuration of the program recommendation device according to the first embodiment, but the operations of some of the functional configuration units and stored data are different in the following points.
The specimen content feature detection unit 205 changes the operation, and also calculates the detection area of the feature when detecting the feature.

The feature importance level calculation unit 206 changes the method of calculating the feature detection amount 501 so that the detection area of each feature content of the feature is used.
FIG. 19 shows a data structure and example contents of the feature importance level information 500 specified by the feature importance level calculation unit 206. As shown in the figure, the feature importance information 500 has the same data configuration as the feature importance information 500 in the first embodiment.

  However, the feature detection amount 501 is not the number of sample content files in which the feature is detected, but the total detection area of all the features detected from each sample content 302 included in the sample content group 300.

The evaluated content feature detection unit 211 changes the operation, and also calculates the detection area of the feature when detecting the feature.
The evaluated content feature storage unit 212 changes the stored data, and stores information on the features detected by the evaluated content feature detection unit 211, including the feature detection area.

The evaluated content evaluation value calculation unit 213 changes the operation, and evaluates the evaluation value 801 by weighting and adding the importance 502 of each feature detected from the evaluated content by the detection area of the feature.
<2-3. Data>
Hereinafter, data changed from the first embodiment will be described.

  The evaluated content feature information 700a of the modified program recommendation apparatus is data obtained by extending the evaluated content feature information 700 of the first embodiment and adding a detection area 2001 corresponding to one-to-one to the features detected from each evaluated content. It is.

Here, the detection area is a relative area indicated by a ratio of the area of the portion where the object corresponding to each feature is detected from the image to the area of the entire image.
FIG. 20 shows the data structure and content example of the evaluated content feature information 700a. As shown in the figure, the evaluated content feature information 700a includes an evaluated content ID 601, a feature ID 401, and a detection area 2001.

  FIG. 21 is an example of the evaluated content evaluation information 800 as a result of the modified program recommendation device that has received the same evaluated content and sample content as in the example of Embodiment 1 evaluating the evaluated content using the detected area. FIG. 22 shows an example of a program recommended based on the result.

  As shown in FIGS. 21 and 22, the evaluation value 801 evaluated by the modified program recommendation device of the second embodiment and the recommended program are the results of evaluation by the program recommendation device of the first embodiment shown in FIGS. Different.

The data generated and updated based on the changed or added data naturally has a different result from the data generated and updated by the program recommendation device of the first embodiment, and will not be shown in the following.
<2-4. Operation>
Hereinafter, processing that is changed or added from the first embodiment will be described.
<2-4-1. Calculation of detection area>
When the modified program recommendation device according to the second embodiment detects a feature in the sample content feature detection unit 205 and the evaluated content feature detection unit 211, it detects whether or not the feature exists in the still image or moving image to be processed. In addition, a detection area indicating how much area the object corresponding to the feature occupies in the processing target image is calculated.

  When a feature is detected from a single still image, local features are extracted from a rectangular area called a detection window, and matching is performed with a feature pattern of an object corresponding to each feature. The area of the detected detection window becomes the area of the object in the still image. The detection area is the area of the object divided by the area of the entire still image.

FIG. 23A shows an example in which two features of a flower and a car are detected from one still image.
In the example shown in the figure, in a still image 2300 having an overall area of 1,000,000, a feature of a flower is detected from a region 2301a having an area of 110,000, and a feature of an automobile is detected from a region 2301b having an area of 270,000. .

In this example, the detection area of the feature of the flower is 110,000 / 1,000,000 = 0.11 and the detection area of the feature of the car is 270,000 / 1,000,000 = 0.27.
The feature detection area in a moving image is obtained by detecting each of the frames selected to detect the feature as a still image and detecting the feature, and calculating the total detection area in the still image in each frame in which the feature is detected. , Divided by the number of selected frames.

  FIG. 23B shows an example in which the feature of a flower is detected from three frames in a moving image. In the example of the figure, the feature of the flower is detected from the region 2301c having the area of 170,000 from the frame 2300a and the flower from the region 2301d having the area of 460,000 is detected from the frame 2300b. And the feature of the flower is not detected from the frame 2300c.

In this example, the detection area in each frame in which the feature of the flower is detected is 0.17 in the frame 2300a and 0.46 in the frame 2300b. Therefore, the detection area in the moving image given in this example is (0.17 + 0.46) /3=0.21 obtained by dividing these sums by the number of frames.
<2-4-2. Processing when obtaining sample content>
The processing for acquiring the sample content is performed in the flow of FIG. 11 as in the first embodiment. However, the contents of the feature detection amount update process (step S1404) performed in this process are changed as follows.

  Hereinafter, a process of updating the feature detection amount 501 with the feature and detection area detected from one sample content by the feature detection amount specifying unit 204 will be described with reference to the flowchart of FIG. Here, the sample content for which the feature is detected is referred to as sample content x.

As shown in the figure, in the feature detection amount update processing in the second embodiment, the detection amount 501 is increased by 1 in the feature detection amount update processing (FIG. 13) in the first embodiment (step S1302). The original detection amount 501 is updated to a value obtained by adding the detection area in the sample content x of the feature (step S1302a), and the others are the same as in the first embodiment.
<2-4-3. Processing when detecting features from evaluated content>
The process for detecting features from the evaluated content is performed according to the flow of FIG. 16 as in the first embodiment. However, some operations have been changed as follows.

Here, as in the case of the description of the first embodiment, the content to be evaluated is referred to as content to be evaluated y.
When the evaluated content feature detection unit 211 detects a feature from the evaluated content y, the detection area 2001 is also calculated for the detected feature (step S1602).

  In the first embodiment, the evaluated content feature detection unit 211 stores the set of the evaluated content ID 601 of the evaluated content y and the detected feature ID 401 in the evaluated content feature storage unit 212. Including the data (step S1603).

The content of the evaluated content y evaluation process (step S1604) performed by the evaluated content evaluation unit 210 is changed as follows.
<2-4-3-1. Evaluated content evaluation process>
In the evaluation processing of the evaluated content performed by the evaluated content evaluation unit 210, the evaluation value 801 is evaluated as the sum of the importance 502 of each feature detected from the evaluated content 602 in the first embodiment. The evaluation value 801 is changed to be evaluated as the sum of products of the importance 502 of each feature detected from the content 602 and the detection area 2001.

  In the following, the evaluated content evaluation processing performed by the evaluated content evaluation unit 210 will be described with reference to the flowchart of FIG. Here, the evaluated content to be evaluated is referred to as evaluated content y.

As shown in the figure, in the evaluated content evaluation processing in the second embodiment, the importance 502 of the selected feature is added to the evaluation value 801 in the evaluated content evaluation processing (FIG. 17) in the first embodiment. As a result, (Step S1703) is changed to add the product of the importance 502 of the selected feature and the detection area 2001 of the feature (Step S1703a), and the others are the same as in the first embodiment.
<3. Embodiment 3>
Next, as the third embodiment, in the program recommendation device 100 according to the first embodiment, the detection amount of each feature has been specified using all the sample contents of the sample content group 300, and the past 7 from the date of evaluation. A modified program recommendation device in which the operation is changed so as to specify the detection amount of each feature using only sample content photographed in a day will be described.

Here, the past 7 days means 7 days including the day of evaluation.
In the following, the description will focus on the part corresponding to the difference from the program recommendation device 100 according to the first embodiment.
<3-1. Overview>
The bar graph of FIG. 26 shows an example of the detected amount of feature A and feature B every day.

In the figure, the horizontal axis indicates the date, and the newer the date is on the right. The right end is the day of evaluation.
The vertical axis indicates the detection amount per day, and the height of the bar indicates the detection amount detected from the sample content photographed on the day.

In this example, the feature A and feature B have the same amount of detection over the entire period, but the amount of detection for seven days surrounded by a dotted line is large for the feature B and small for the feature A.
From this, it is considered that feature B is highly interested in the last seven days and is still highly interested. On the other hand, the feature A was once highly interested, but is considered to be less interested now.

The modified program recommendation device according to the third embodiment evaluates the content to be evaluated based on the detected amount of each feature specified only by the sample content photographed in the past seven days from the date of evaluation, so that the user's interest As content changes from day to day, we recommend content that is considered to be of high interest.
<3-2. Configuration>
The hardware configuration of the modified program recommendation device is exactly the same as the hardware configuration of the program recommendation device according to the first embodiment.

The functional configuration of the modified program recommendation device is also substantially the same as the functional configuration of the program recommendation device according to the first embodiment, but the operations of some of the functional configuration units and stored data are different in the following points.
The sample content storage unit 202 changes the content of the stored data, and stores the sample content group 300a including the data of the photographing time 2701 added in advance to each sample content in the sample content group 300.

  FIG. 27 shows a data configuration and example contents of the sample content group 300a in the third embodiment. As shown in the figure, the data of the photographing time 2701 corresponding to each sample content is added to the configuration of the sample content group 300 of the first embodiment.

As the photographing time, for example, information added in advance to the specimen content as Exif (Exchangeable image file format) information is used.
The sample content feature detection unit changes the operation so that the feature is detected only from the sample content whose shooting time is within the range of the past seven days. Details of the changed operation will be described later.
<3-3. Operation>
Hereinafter, processing that is changed or added from the first embodiment will be described.

Next, as shown in FIG. 28, the process for acquiring the sample content is changed to a process for acquiring and storing all the sample contents to be acquired.
In this process, the process (step S1101) for determining whether or not there is sample content to be acquired from the process (FIG. 11) when acquiring sample content in the first embodiment and one sample content to be acquired are acquired. This excludes processing (step S1102).

  When the sample content 302 acquired by the sample content acquisition unit 201 is stored in the sample content storage unit 202, the sample content acquisition unit 201 also stores data of the photographing time 2701 of the sample content.

  As shown in FIG. 29, the process for detecting a feature from the evaluated content is changed to a process that attempts to detect a feature from each of the evaluated content that has not been detected and records the detected feature. Is done.

This process is obtained by omitting the evaluation process (step S1604) of the evaluated content from the process (FIG. 16) for detecting features from the evaluated content in the first embodiment.
As shown in FIG. 30, in the process of recommending a program, in the process of recommending a program in the first embodiment (FIG. 18), the importance of all the features is specified before the process of selecting the evaluated content (step S1802). The processing is changed so as to perform the processing (step S3001) and the evaluation processing of all evaluated content (step S3002).

  Here, the evaluation process for all evaluated content (step S3002) is a process in which the evaluated content evaluation value calculation unit 213 evaluates all the evaluated content. This process is equivalent to the evaluation process for the evaluated content group shown in FIG.

The feature level identification processing (step S3001) is performed by the feature detection amount identification unit 204. The contents of this processing will be described below using the flowchart of FIG.
First, the detection amounts of all features are set to 0 (step S3101).

  Next, it is determined whether or not there is sample content in the sample content group 300 that has not been selected since the start of the current specific processing (step S3102). If there is any (step S3102: YES), the following processing is performed.

  One sample content that has not been selected is selected from the sample content group 300 (step S3103). Hereinafter, the sample content selected here will be referred to as sample content x.

  Next, it is determined whether or not the photographing time 2701 of the sample content x is within the range of the past seven days (step S3104). If there is, the following processing is performed (step S3104: Yes). If not, the process returns to step S3102 (step S3104: No).

An attempt is made to detect the feature from the sample content x (step S3105). This process is the same as the feature detection process in the first embodiment.
Then, the detection amount of the feature detected from the sample content x is updated (step S3106). This processing is the same as the feature detection amount update processing (FIG. 13) in the first embodiment.

When the feature detection amount update processing ends, the process returns to step S3102.
When there is no sample content in the sample content group 300 that has not yet been selected since the process of specifying the importance levels of all the features is started (step S3102: No), the importance levels of the features are calculated. This processing is the same as the feature importance calculation processing (FIG. 14) in the first embodiment.
<4. Embodiment 4>
Next, as Embodiment 4, in the program recommendation device 100 according to Embodiment 1, the evaluated content is evaluated according to the value of the feature detection amount, and the change in the feature detection amount in the past seven days is further changed. A modified program recommendation device whose operation is changed so as to evaluate the evaluated content will be described.

In the following, the description will focus on the part corresponding to the difference from the program recommendation device 100 according to the first embodiment.
<4-1. Overview>
The bar graph in FIG. 32 shows an example of the detected amount of feature A and feature B every day.

In the figure, the horizontal axis indicates the date, and the newer the date is on the right. The right end is the day of evaluation.
The vertical axis indicates the detection amount per day, and the height of the bar indicates the detection amount detected from the sample content photographed on the day.

  In this example, feature A and feature B have the same total detection amount in the period surrounded by the dotted line, but when looking at the amount of change in the detection amount from 7 days ago to the current day, feature B is the detection amount. Has increased significantly.

From this, it can be considered that the feature B is currently very much interested.
On the other hand, it is conceivable that a feature with a large detection amount and a small change is a feature that is easy to be reflected rather than having a high interest of the user.

  The modified program recommendation device according to the fourth embodiment is specified by the sample content photographed seven days before the detection amount specified by the sample content photographed on the day of evaluation in addition to the detection amount from the entire period. By evaluating the content to be evaluated using the amount of change from the detected amount, the content to be evaluated that is considered to be of high interest is recommended as the user's interest varies from day to day.

In the following description, one day of the evaluation day is referred to as a first period, and one day seven days before the evaluation day is referred to as a second period.
<4-2. Configuration>
The hardware configuration of the modified program recommendation device is exactly the same as the hardware configuration of the program recommendation device according to the first embodiment.

The functional configuration of the modified program recommendation device is also substantially the same as the functional configuration of the program recommendation device according to the first embodiment, but the operations of some of the functional configuration units and stored data are different in the following points.
The sample content storage unit 202 changes the content of the stored data, and stores data having the same configuration as the sample content group 300a (FIG. 27) in the third embodiment.

  The feature importance calculation unit 206 changes the operation and specifies the detection amount and the importance corresponding to the first period and the second period in addition to the same detection amount and importance as those used in the first embodiment. I am doing so.

  Also, the importance calculation method of each feature used by the evaluated content evaluation unit 210 is changed so that the detected amount of each feature is divided by the maximum detected amount of all the features, and the change in the importance of the feature is further changed. The importance is calculated by multiplying the quantity. Here, the amount of change in the importance of the feature is a value obtained by subtracting the importance corresponding to the second period from the importance corresponding to the first period of the feature.

The feature importance storage unit 207 changes the content of the stored data, and stores feature importance information 500a including the detection amount and the importance for each of the first period and the second period.
FIG. 33 shows the data structure and example contents of the feature importance information 500a in the fourth embodiment. As shown in the figure, in addition to the configuration of the feature importance information 500 in the first embodiment, the feature importance information 500a includes the first period detection amount information 3301 and the second period detection amount corresponding to each feature. It consists of information 3302.

Furthermore, the detection amount information 3301 in the first period and the detection amount information 3302 in the second period are configured by detection amounts and importance levels corresponding to the respective detection amounts.
Hereinafter, the detection amount 501a and the importance level 502a included in the detection amount information 3301 in the first period are referred to as a first detection amount 501a and a first importance level 502a, respectively. The included detection amount 501b and importance 502b are referred to as a second detection amount 501b and a second importance 502b, respectively.

  Further, the detection amount 501 and importance 502 common to the feature importance information 500 in the first embodiment are distinguished from the first and second detection amounts and importance in order to distinguish between the detection amount 501 and importance of the entire period. It is assumed that 502.

In the example of FIG. 33, since the importance of the feature of FE0001 increases from the second period to the first period, the importance of all periods is a positive value. On the other hand, the feature of FE0004 always has a large amount of detection, but the importance does not increase, so the importance of the whole period is zero. In addition, the feature of FE0002 whose importance is lowered has a negative importance for the entire period.
<4-3. Operation>
Hereinafter, processing that is changed or added from the first embodiment will be described.

The process for acquiring the sample content is performed in the same manner as the process for acquiring the sample content in the third embodiment (FIG. 28).
The process for detecting features from the evaluated content is performed in the same manner as the process for detecting features from the evaluated content in the third embodiment (FIG. 29).

  The process of recommending a program is performed in the same manner as the process of recommending a program in the third embodiment (FIG. 30). However, since the process of specifying the importance of all features (step S3001) is different from that of the third embodiment, it will be described in detail below.

The identification processing of importance levels of all the features in the fourth embodiment is obtained by changing a part of the same processing (FIG. 31) in the third embodiment.
In the following, this process will be described with reference to the flowchart of FIG. Here, the sample content selected in step S3103 is referred to as sample content x.

  First, the process of initializing the detection amount 501 of all features to 0 in the third embodiment (step S3101) sets the detection amount 501, the first detection amount 501a, and the second detection amount 501b of all features to 0. It replaces with the processing (step S3101a) to initialize.

The determination based on the shooting time in the third embodiment (step S3104) is not performed.
In addition, the feature detection amount update process (step S3106) and the feature importance calculation process (step S3107) are partially changed, and will be described below.
<4-3-1. Update processing of feature detection amount>
The update processing of the feature detection amount in the fourth embodiment is a part of the same processing (FIG. 13) in the first and third embodiments, and the detection amount 501 of the entire feature period is updated. Further, the first detection amount 501a or the second detection amount 501b of the feature is further updated as necessary.

  In the following, changes in this process will be described using the flowchart of FIG. Here, the sample content to be processed is referred to as sample content x. In addition, a feature that is a target for updating each detection amount after step S1302 is referred to as a feature z.

  In the feature detection amount update process according to the fourth embodiment, after the process of recording the detection amount 501 for the entire period (step S1303), the determination of the shooting time (step S3501) and the process of updating the first detection amount 501a ( Steps S1302b and S1303b) and processing for updating the second detection amount 501b (steps S1302c and S1303c) are added.

It is determined whether the photographing time 2701 of the sample content x enters the first period or the second period (step S3501).
When the photographing time 2701 of the sample content x enters the first period (step S3501: within the first period), the value of the first detection amount 501a of the feature z is set from the original first detection amount 501a. The value is updated by 1 (step S1302b).

Then, the updated first detection amount 501a of the feature z is stored in the feature importance storage unit 207 (step S1303b).
When the update and storage of the first detection amount 501a are completed, the process returns to step S1301.

  When the imaging time 2701 of the sample content x enters the second period (step S3501: within the second period), the second update of the first detection amount 501a and the storage process of the feature z are performed in the same manner as the above-described processing for updating and storing the first detection amount 501a. 2 detection amount 501b is updated (step S1302c), stored in the feature importance storage unit 207 (step S1303c), and the process returns to step S1501.

When the photographing time 2701 of the sample content x does not enter either the first period or the second period (step S3501: out of period), the process directly returns to step S1501.
<4-3-2. Feature importance calculation processing>
The feature importance calculation processing in the fourth embodiment is obtained by changing a part of the same processing in the first and third embodiments. In the first and third embodiments, the importance level 502 of the whole period of the feature is calculated. The first importance level 502a and the second importance level 502b of the feature are calculated, and the importance level 502 of the whole period is calculated based on the result. It is changed so as to calculate.

In the following, this process will be described with reference to the flowchart of FIG.
In the feature importance calculation processing according to the fourth embodiment, the first importance 502a is calculated and recorded (step S1402a) before the processing (step S1402 and step S1403) for calculating and recording the importance 502 for the entire period. And step S1403a) and a process for calculating and recording the second importance 502b (step S1402b and step S1403b), and a process for calculating the importance 502 for the entire period (step S1402) is changed. (Step S1402c).

In the following, a process for calculating and recording the importance of the feature z will be described, assuming that one feature to be processed and calculated for the importance is called a feature z.
First, the value of the first importance 502a of the feature z is calculated as a value obtained by dividing the detected amount 501a of the feature z by the largest first detected amount 501a among the first detected amounts 501a of all the features (step S1402a).

Then, the first importance 502a of the calculated feature z is stored in the feature importance storage unit 207 (step S1403a).
Similarly, the second importance 502b of the feature z is also calculated (step S1402b) and stored in the feature importance storage unit 207 (step S1403b).

  When the calculation of the first importance 502a and the second importance 502b of the feature z is completed, the importance 502 for the entire period is calculated (step S1402c). The importance level 502 of the whole period of the feature z is obtained by dividing the detection amount 501 of the whole period of the feature z by the detection amount 501 of the largest whole period among the detection quantity 501 of the whole period. It is calculated by multiplying the difference obtained by subtracting the importance of 2.

When step S1403 ends, the process returns to step S1401.
<5. Embodiment 5>
Next, as a fifth embodiment, the program recommending apparatus 100 according to the first embodiment operates so as to detect the feature detected from the evaluated content by image processing from the metadata previously added to the evaluated content. A modified program recommendation device in which is changed will be described.
<5-1. Overview>
Since the sample content is a photograph or video that the user owns, metadata is not necessarily added. Even if it is added, the metadata appropriately indicates the characteristics of the sample content. There are few.

  Therefore, in order to evaluate the content to be evaluated using the metadata of the sample content, it is necessary for the user to add appropriate metadata to the sample content. However, this method takes time and effort for the user.

On the other hand, the content to be evaluated is a broadcast program, and it can be expected that the broadcaster has added appropriate metadata in advance for convenience such as searching.
Therefore, if metadata is used when detecting features from the content to be evaluated, it is possible to evaluate the content to be evaluated at high speed by comparison processing of simple text or the like without using complicated image processing. Conceivable.

The modified program recommendation device according to the fifth embodiment detects features from the sample content by image processing as in the first embodiment, and detects features from the evaluated content by metadata.
<5-2. Configuration>
The hardware configuration of the modified program recommendation device is exactly the same as the hardware configuration of the program recommendation device according to the first embodiment.

The functional configuration of the modified program recommendation device is also substantially the same as the functional configuration of the program recommendation device according to the first embodiment, but the operations of some of the functional configuration units and stored data are different in the following points.
The content-to-be-evaluated storage unit 209 changes the content of the data to be stored, and stores the content to be evaluated by adding metadata corresponding to each content to be evaluated.

  FIG. 37 shows a data structure and an example of contents of the evaluated content group 600a stored in the evaluated content storage unit 209. As shown in the figure, it is data obtained by adding metadata 3701 of a character string corresponding to each evaluated content to the evaluated content group 600 (see FIG. 6) of the first embodiment.

  The metadata 3701 is data added in advance to each content to be evaluated, and here is a character string of at least one keyword related to the content of the content to be evaluated.

In the example of FIG. 37, metadata of character strings “flower” and “mountain” is added to the evaluated content 602 of EC0001.
The feature detection reference storage unit 203 changes the content of stored data, and includes feature detection including a reference 3801 for detecting features from the metadata 3701 of the evaluated content 602 and a reference 3802 for detecting features from the data of the sample content 302. Reference information 400a is stored.

  Hereinafter, the criterion 3801 for detecting features from the metadata of the evaluated content is referred to as the criterion 3801 of evaluated content, and the reference 3802 of detecting the features from the sample content data is referred to as the sample content reference 3802.

  FIG. 38 shows an example of the data structure and contents of the feature detection reference information 400a. As shown in the figure, the feature detection criterion information 400a is composed of a feature ID 401 and a feature detection criterion 402a, and the feature detection criterion 402a is composed of an evaluation content criterion 3801 and a sample content criterion 3802.

  The evaluated content standard 3801 is a character string corresponding to a keyword included in the metadata 3701 of the evaluated content 602. In the example of FIG. 38, the standard 3801 of the evaluated content of the feature of FE0001 is a character string “automobile”, and if the metadata 3701 of the evaluated content 602 includes the keyword “automobile”, the feature of FE0001 Can be detected from the evaluated content 602.

  The sample content reference 3802 is a reference by which the sample content feature detection unit 205 detects a feature from the sample content, and is data corresponding to the feature detection reference 402 of the feature detection reference information 400 in the first embodiment. In the example of FIG. 38, the reference 3802 of the sample content of the feature of FE0001 is feature amount information corresponding to the car image.

  The sample content feature detection unit 205 changes the operation, and detects the feature from the sample content 302 data according to the sample content reference of the feature detection reference information. The details of the operation are not described here because the processing itself is not changed only by changing the reference used for feature detection only for the sample content.

The evaluated content feature detection unit 211 changes the operation, and detects a feature from the metadata of the evaluated content according to the criterion of the evaluated content in the feature detection reference information. Details of the changed operation will be described later.
<5-3. Operation>
Below, the process which tries detection of the characteristic from the to-be-evaluated content changed from Embodiment 1 is demonstrated.

  A process when the evaluated content feature detection unit 211 tries to detect a certain feature from the evaluated content 602 is described below. Here, the evaluated content to be processed is called the evaluated content y, and the feature to be detected is called the feature z.

First, the metadata 3701 added to the evaluated content y is acquired from the evaluated content group 600 a stored in the evaluated content storage unit 209.
Next, from the feature detection reference information 400a stored in the feature detection reference storage unit 203, the reference 3801 of the content to be evaluated for the feature z is acquired.

  Then, it is determined whether the metadata 3701 of the evaluated content y includes a keyword that matches the criterion 3801 of the evaluated content of the feature z. If it is included, the feature z is detected from the evaluated content y, and if it is not included, the feature z is not detected from the evaluated content y.

  Here, the determination as to whether or not a matching keyword is included is made by comparing the character string of the criterion 3801 of the evaluated content with the characteristic z with each keyword included in the metadata 3701 of the evaluated content y.

  Taking FIG. 37 and FIG. 38 as an example, since the metadata 3701 including the keyword “flower” is added to the evaluated content of EC0001, the feature of FE0004 having the criterion 3801 of the evaluated content “flower”. Is detected.

On the other hand, if the keyword “kimono” is not included in the metadata 3701 of the evaluated content of EC0001, the feature of the FE0003 having the criterion 3801 of the evaluated content “kimono” is not detected.
<Supplement>
Although the content evaluation apparatus according to the present invention has been described based on the embodiment, the present invention is not limited to the program recommendation apparatus as shown in the above-described embodiment.

  (1) The present invention performs the evaluation processing of the content to be evaluated shown in the embodiment, the processing for specifying the feature detection amount using the sample content, and the like (see FIGS. 11 and 13 to 18), etc. In addition, a control program including program codes to be executed by various circuits connected to the processor can be recorded on a recording medium or distributed and distributed via various communication paths. Examples of such a recording medium include an IC card, a hard disk, an optical disk, a flexible disk, and a ROM. The distributed and distributed control program is used by being stored in a memory or the like that can be read by the processor, and the processor executes the control program to realize the functions shown in the embodiments. Become so. A part of the control program is transmitted to a device (processor) capable of executing a program separate from the content evaluation device via various networks, and the control program is partially executed by the device capable of executing the program. It is also possible to make it.

  (2) Part or all of the constituent elements constituting the content evaluation apparatus may be mounted as one or a plurality of integrated circuits (IC, LSI, etc.), and other elements are included in the constituent elements of the content evaluation apparatus. It is also possible to add an integrated circuit (single chip).

  (3) In the first to fifth embodiments, the content evaluation device recommends the content to be evaluated. However, the content evaluation device is not limited to the recommended device as long as the content to be evaluated is evaluated. For example, a recording device having a function of deleting evaluated content with a low evaluation value from the evaluated content storage unit may be used, or a device that displays an evaluation value of each evaluated content. Moreover, the apparatus which stores the to-be-evaluated content which added the evaluation value as metadata to a recording medium may be sufficient.

  (4) In the first to fifth embodiments, the content to be evaluated is broadcast by wire or wireless. However, the content to be evaluated is not limited to be broadcast as long as the content evaluation device can acquire the content. For example, the evaluated content may be downloaded from the Internet, or may be recorded on various recording media such as a hard disk.

  (5) In the first to fifth embodiments, the content to be evaluated is a moving image, but it is not limited to a moving image as long as the feature can be detected. For example, the content to be evaluated may be a still image, music, or sound.

When the content to be evaluated is music or speech, it is possible to detect features such as words contained in the music or speech using speech recognition technology.
(6) In Embodiments 1 to 5, the sample content is stored in a recording medium such as a memory card, but is not limited to a memory card. For example, it may be stored in a hard disk or a flexible disk.

Moreover, it is not necessarily stored in the recording medium. For example, it may be downloaded to a content evaluation apparatus through a network.
In short, the sample content may be in any form as long as it can be acquired by the content evaluation apparatus.

  (7) In Embodiments 1 to 5, the sample content is a still image and a moving image. However, the sample content is not limited to a still image and a moving image as long as the feature can be detected. For example, the content to be evaluated may be music or voice. Moreover, it is not necessary to be both a still image and a moving image, and either one can be used.

  (8) In the first to fifth embodiments, the feature indicates the content that the object is reflected in the image, and an example using the object detection technique is given. However, if there is a reference that can be detected by the content of the sample content, The feature is not limited to indicating that the object is shown.

For example, a result obtained by classifying feature amounts extracted from the entire image may be used as a feature.
Further, it may be a feature that the average of the luminance values of all the pixels in the image is a predetermined value or more, or a feature that can be detected from the frequency distribution of each pixel in the image.

In the case where the sample content is music or voice, a feature that a predetermined word is included may be used.
(9) In Embodiments 1 to 5, the output destination is an external display device. However, if the output is based on the evaluation result of the content to be evaluated, the display is not always displayed on the external display device. Absent. For example, the result may be displayed on a display screen built in the main body, or may be output to a remote computer connected via the Internet.

  (10) In the first to fifth embodiments, the recommendation is made with the operation by the controller which is an external device as an opportunity, but the operation is not limited to the operation by the controller by the outside. For example, it is good also as operating with the button with which an apparatus main body is provided, and it is good also as receiving operation using the touchscreen on a display screen.

Further, it is not always necessary to make a recommendation in response to an operation, and the content to be evaluated may be automatically recommended without an explicit operation.
(11) In the first to fifth embodiments, the content evaluation apparatus includes a memory as hardware for storing data and a program. However, as long as data and a program can be stored, the content evaluation apparatus is not limited to the memory. For example, it may be a hard disk or a flexible disk.

  (12) In the first to fifth embodiments, the sample content acquisition unit generates the sample content ID and assigns the sample content ID to each sample content. However, if each sample content can be uniquely identified in the own device, the sample content is not necessarily provided. There is no need to use a content ID. For example, if the file name of the sample content can be uniquely identified, the file name may be used instead of the sample content ID.

  Also, when the sample content ID is used, the sample content ID is not limited to an alphabetic character added to the beginning of the serial number, and may be a random character string or a hash value of sample content data.

  Similarly, the feature ID and the content to be evaluated ID are not limited to those having the alphabet at the beginning of the serial number as long as the feature and the content to be evaluated can be uniquely identified within the own device.

  (13) In the first to fifth embodiments, the file is acquired as the sample content. However, if the sample content can be distinguished one by one, it is not necessary to acquire the file. For example, each piece of data included in an archive file obtained by collecting a plurality of data into one file may be used as the sample content.

  (14) In the first to fifth embodiments, the feature detection reference information is predetermined and does not change. However, the feature detection reference information may change dynamically. For example, feature quantities such as SIFT feature quantities may be extracted from each sample content, classified using the data classification technique based on the extracted feature quantities, and feature detection reference information may be determined based on the classification result. Further, the feature detection reference information may be acquired from the Internet or various recording media.

  (15) In the first to fifth embodiments, the importance of each feature is a value obtained by dividing the detection amount by the maximum value of the detection amount of all the features. However, if based on the detection amount, the detection amount is the detection amount of all the features. It is not limited to the value divided by the maximum value of. For example, a value obtained by dividing the detection amount by the total number of sample contents may be used, or the detection amount may be used as the importance as it is.

  (16) In the first to fifth embodiments, the evaluation value of the evaluated content is the sum of the importance calculated from the detected amount of each feature detected from the evaluated content, the detected amount of the first period, and the second Although the sum of the differences from the importance of the period is set as the sum, the evaluation is not limited to this as long as the evaluation is based on the feature detection amount. For example, the detection amount of each feature may be added as it is, or the product of the importance of each feature may be used. In addition, the evaluation may be performed in consideration of not only the feature detection amount but also other criteria.

  (17) In the first to fifth embodiments, the SIFT feature amount and the color information are exemplified as the feature detection reference. However, the feature detection reference is not limited thereto as long as the feature can be detected from the sample content or the evaluated content. For example, it may be information on a feature quantity that can be extracted based on another standard such as a Gabor feature quantity, or may be data of a program that performs feature detection.

  (18) In the first to fifth embodiments, it is assumed that the processing for acquiring the sample content is executed when the memory card is inserted into the own device, but the memory card is inserted into the own device when the processing is performed. It is not necessary to be limited to when. For example, it is good also as performing the process at the time of acquiring sample content triggered by a user's operation.

  (19) In the first to fifth embodiments, the process for detecting the feature from the evaluated content is periodically performed. However, it is not always necessary to periodically perform the process. For example, a process for detecting a feature from the evaluated content may be executed in response to a user operation, or the process may be executed every time the evaluated content is acquired.

  (20) In the first to fifth embodiments, a predetermined number of content to be evaluated is output based on the result of evaluating the content to be evaluated. However, if output based on the evaluation result is performed, a predetermined number of content to be evaluated is output. The content to be evaluated is not necessarily output. For example, all content to be evaluated may be output in order from the highest evaluation value.

  (20) In the second embodiment, the detection area calculated from the area of the detection window used for detecting the feature is used as the value indicating the degree to which the sample content and the evaluated content include the feature. However, the present invention is not limited to this. . For example, if the feature that the average of the luminance values of all the pixels in the image is greater than or equal to a predetermined value is used, the average of the luminance values can be used as a value indicating the degree of including the feature.

  (21) In the third embodiment, the sample content photographed in the past seven days from the date of evaluation is used, but the period is not limited to the past seven days from the date of evaluation. For example, it may be the past month from the date of evaluation, or may be a predetermined number of days in the past from the date when the latest sample content was taken.

In addition, the shooting date and time stored as Exif information is used, but the file creation date and update date and time may be used.
(21) In the fourth embodiment, the amount of change in the detection amount is obtained based on the difference between the importance level of each feature in the first period and the importance level in the second period. For example, a value obtained by dividing the importance of the first period by the importance of the second period may be used, or may be obtained by a difference between the detection amount of the first period and the detection amount of the second period. Also good.

  (22) In the fourth embodiment, the importance of each feature is obtained by multiplying the detection amount from the entire period by the change amount of the detection amount, but it is not necessary to limit to the entire period. For example, as in the third embodiment, the detection amount may be specified from the sample content photographed in a predetermined period such as the past seven days, and the detection amount may be multiplied by the change amount of the detection amount.

  Further, although the detection amount from the entire period is multiplied by the change amount of the detection amount, it is not necessary to be limited to multiplication. For example, the importance of each feature may be obtained by the sum of the detection amount from the entire period and the change amount of the detection amount, or the detection amount change amount itself is used as the importance without using the detection amount from the entire period. Also good.

  Furthermore, the amount of change in the detection amount may be a negative value or a value larger than 1 depending on the calculation method, but this is normalized so that it falls within the range of 0 to 1, and the importance of each feature is determined. It may be used for calculation. For example, when the amount of change in the detected amount becomes a negative number, 0 may be used instead of the amount of change in the detected amount.

  (23) In the fifth embodiment, the feature is detected from the evaluated content using the keyword text included in the metadata. However, the feature need not be limited to the text as long as the metadata feature can be detected. For example, the feature may be detected by a numerical value representing a genre or the like.

  The content evaluation apparatus and method according to the present invention are useful as devices for recording broadcast programs. The present invention can also be applied to uses such as a home server that holds personal contents and a service providing server that provides services on a network.

DESCRIPTION OF SYMBOLS 100 Program recommendation apparatus 101 Broadcast 102 Display apparatus 103 Digital camera 104 Memory card 105 Controller 201 Sample content acquisition part 202 Sample content storage part 203 Feature detection reference | standard storage part 204 Feature detection amount specific | specification part 205 Sample content feature detection part 206 Feature importance calculation Unit 207 Feature importance storage unit 208 Evaluated content acquisition unit 209 Evaluated content storage unit 210 Evaluated content evaluation unit 211 Evaluated content feature detection unit 212 Evaluated content feature storage unit 213 Evaluated content evaluation value calculation unit 214 Evaluated Content evaluation value storage unit 215 Operation input unit 216 Content selection unit 217 Output unit 300 Sample content group 301 Sample content ID
302 Specimen content 400 Feature detection reference information 401 Feature ID
402 Feature Detection Criteria 500 Feature Importance Information 501 Detection Amount 502 Importance 600 Evaluated Content Group 601 Evaluated Content ID
602 Evaluated content 700 Evaluated content feature information 800 Evaluated content evaluation information 801 Evaluation value 2001 Detection area 2701 Shooting time 3301 Detection amount information in the first period 3302 Detection amount information in the second period 3701 Metadata 3801 Evaluated content Feature detection criteria 3802 Sample content feature detection criteria

Claims (16)

A content evaluation apparatus for evaluating a plurality of evaluated contents,
Sample content acquisition means for acquiring a plurality of sample contents used for evaluating the evaluated content from among a plurality of contents owned by the user ;
From the sample content of all acquired or part by the sample content obtaining unit, attempts a plurality of features each detection, for each detected feature, identifying the detected amount is the amount of sample content that the feature has been detected A feature detection amount specifying means;
Oite to one of the evaluation contents for each feature which the evaluation content of the one having one of the plurality of features, of the evaluation content of the one based on the detected amount specified by the feature detection amount specifying means And an evaluated content evaluation means for performing evaluation ,
The feature detection amount specifying means calculates importance for each detected feature based on a detection amount corresponding to the feature and indicating how much is detected from the sample content,
The evaluated content evaluation unit calculates each of the plurality of evaluated contents by trying to detect each of the plurality of features from the evaluated content and adding the importance corresponding to each detected feature. The evaluation is performed based on the evaluation value of the evaluated content, and a predetermined number of the evaluated content is selected in descending order of the evaluation value.
A content evaluation apparatus characterized by that. The feature detection amount specifying means determines a feature detection amount.
The content evaluation apparatus according to claim 1, wherein each sample content in which the feature is detected is calculated based on a sum of values indicating a degree to which the feature is included. The sample content is a still image, and each feature represents each object included in the still image,
The feature detection amount specifying means includes:
A value indicating the degree to which one sample content in which one feature is detected includes the feature,
The content evaluation apparatus according to claim 2, wherein the area of a portion where the object represented by the feature is included in the sample content is calculated as a ratio with respect to the area of the sample content. The sample content is a moving image, and each feature represents each object included in a frame in the moving image.
The feature detection amount specifying means tries to detect each feature using a specific type of frame as a detection target,
For one sample content where one feature is detected,
For each frame containing the object represented by the feature of the sample content,
Calculate the ratio of the area of the frame containing the object to the area of the frame,
A value indicating the degree to which the sample content includes the feature,
The content evaluation apparatus according to claim 2, wherein a ratio sum of the ratios of each frame including the object of the sample content is calculated as a ratio to the number of frames that are the detection target of the sample content. The sample content is a moving image, and each feature represents each object included in a frame in the moving image.
The feature detection amount specifying means tries to detect each feature using a specific type of frame as a detection target,
A value indicating the degree to which one sample content in which one feature is detected includes the feature,
The number of frames in which the object represented by the feature of the sample content is included,
The content evaluation apparatus according to claim 2, wherein the content is calculated as a ratio of the sample content to the number of frames to be detected. The feature detection amount specifying means determines a feature detection amount.
The content evaluation apparatus according to claim 1, wherein the content evaluation device calculates the feature based on the number of sample contents in which the feature is detected. The evaluated content evaluation means includes:
The sum of the importance values calculated from the detected amount of each feature corresponding to one evaluated content
The content evaluation apparatus according to claim 1, wherein the evaluation of the content to be evaluated is performed based on the content. The sample content is content to which information on the time of creation, acquisition or update is given,
The feature detection amount specifying means includes:
2. The content evaluation apparatus according to claim 1, wherein the given time tries to detect a feature from sample content that falls within a range from a time when the evaluated content evaluation unit performs evaluation to a predetermined period before. . The sample content is content to which information on the time of creation, acquisition or update is given,
The feature detection amount specifying means includes:
Sample content that falls within a first period in which the given time is in a range from the time when the evaluated content evaluation means performs evaluation to a predetermined period before,
Attempts to detect features from sample content that falls within a second period in which the given time is in a predetermined range past the first period,
The evaluated content evaluation means includes a feature corresponding to one evaluated content,
The degree of importance is calculated by using, instead of the detection amount, the amount of change of the detection amount specified by the sample content whose time is in the first period with respect to the detection amount specified by the sample content whose time is in the second period content evaluation apparatus according to claim 1, characterized in that. The sample content is content to which information on the time of creation, acquisition or update is given,
The feature detection amount specifying means includes:
The feature time is tried to detect a feature from the sample content that falls within a range from a newest time among the times given to the sample content to a predetermined period before. Content evaluation device. The evaluated content is content to which metadata is added,
The content evaluation device includes:
A means for associating the metadata of the evaluated content with the characteristics;
The evaluated content evaluation means includes:
The content evaluation apparatus according to claim 1, wherein the evaluation of the evaluated content is performed using the importance calculated based on the detected amount of each feature corresponding to the metadata of the evaluated content. . The evaluated content evaluation means performs the evaluation on a plurality of evaluated contents,
The content evaluation device includes:
Wherein the each of the evaluation content that is evaluated by the evaluation content evaluation unit, the content evaluation apparatus according to claim 1, further comprising a display means for displaying the order from the high not be evaluated contents of the evaluation value. A content evaluation method executed by a content evaluation apparatus that includes a sample content acquisition unit, a feature amount detection amount specifying unit, and an evaluated content evaluation unit, and that evaluates a plurality of evaluated content,
The sample content acquisition unit acquires a plurality of sample contents used for evaluating the evaluated content from among a plurality of contents owned by the user ;
The feature detection amount specifying unit, from the specimen content of the sample content obtaining all acquired in step or part, attempted a plurality of features each detection, for each detected feature, the specimen content the feature has been detected A feature detection amount specifying step for specifying a detection amount that is a quantity;
Wherein the evaluation content evaluation unit, in one of the evaluation contents for each feature which the evaluation content of the one of the plurality of features has, the based on the detected amount specified by the feature detection amount specifying step and a evaluation target content evaluation step of evaluating one of the evaluation content,
The feature detection amount specifying step calculates, for each of the detected features, an importance indicating a ratio of how much is detected from the sample content based on a detection amount corresponding to the feature,
The evaluated content evaluation step calculates each of the plurality of evaluated contents by trying to detect each of the plurality of features from the evaluated content and adding the importance corresponding to each detected feature. A content evaluation method comprising: performing the evaluation based on an evaluation value of the evaluated content, and selecting a predetermined number of evaluated content in descending order of the evaluation value . A content evaluation program executed by a content evaluation apparatus that evaluates a plurality of evaluated contents,
In the content evaluation device,
A sample content acquisition step for acquiring a plurality of sample contents used for evaluation of the evaluated content from among a plurality of contents owned by the user ;
From the sample content of the sample content obtaining all acquired in step or part, attempted a plurality of features each detection, for each detected feature, identifying the detected amount is the amount of sample content that the feature has been detected A feature detection amount identification step;
In one of the evaluation contents for each feature which the evaluation content of the one of the plurality of features having the evaluation of the evaluation content of the one based on the detected amount specified by the feature detection amount specifying step Performing the evaluated content evaluation step to be performed ,
The feature detection amount specifying step calculates, for each of the detected features, an importance indicating a ratio of how much is detected from the sample content based on a detection amount corresponding to the feature,
The evaluated content evaluation step calculates each of the plurality of evaluated contents by trying to detect each of the plurality of features from the evaluated content and adding the importance corresponding to each detected feature. A content evaluation program characterized in that the evaluation is performed based on an evaluation value of the evaluated content, and a predetermined number of evaluated content is selected in descending order of the evaluation value . A recording medium for recording a content evaluation program executed by a content evaluation apparatus that evaluates a plurality of evaluated contents,
The content evaluation program includes:
In the content evaluation device,
A sample content acquisition step for acquiring a plurality of sample contents used for evaluation of the evaluated content from among a plurality of contents owned by the user ;
From the sample content of the sample content obtaining all acquired in step or part, attempted a plurality of features each detection, for each detected feature, identifying the detected amount is the amount of sample content that the feature has been detected A feature detection amount identification step;
In one of the evaluation contents for each feature which the evaluation content of the one of the plurality of features having the evaluation of the evaluation content of the one based on the detected amount specified by the feature detection amount specifying step Performing the evaluated content evaluation step to be performed ,
The feature detection amount specifying step calculates, for each of the detected features, an importance indicating a ratio of how much is detected from the sample content based on a detection amount corresponding to the feature,
The evaluated content evaluation step calculates each of the plurality of evaluated contents by trying to detect each of the plurality of features from the evaluated content and adding the importance corresponding to each detected feature. A recording medium , wherein the evaluation is performed based on an evaluation value of the evaluated content, and a predetermined number of the evaluated content is selected in descending order of the evaluation value . An integrated circuit constituting a content evaluation apparatus for evaluating a plurality of evaluated contents,
Sample content acquisition means for acquiring a plurality of sample contents used for evaluating the evaluated content from among a plurality of contents owned by the user ;
From the sample content of all acquired or part by the sample content obtaining unit, attempts a plurality of features each detection, for each detected feature, identifying the detected amount is the amount of sample content that the feature has been detected A feature detection amount specifying means;
Oite to one of the evaluation contents for each feature which the evaluation content of the one having one of the plurality of features, of the evaluation content of the one based on the detected amount specified by the feature detection amount specifying means And an evaluated content evaluation means for performing evaluation ,
The feature detection amount specifying means calculates importance for each detected feature based on a detection amount corresponding to the feature and indicating how much is detected from the sample content,
The evaluated content evaluation unit calculates each of the plurality of evaluated contents by trying to detect each of the plurality of features from the evaluated content and adding the importance corresponding to each detected feature. An integrated circuit , wherein the evaluation is performed based on an evaluation value of the evaluated content, and a predetermined number of the evaluated content is selected in descending order of the evaluation value .

Images