JP2002063381A - Registration system for multimedia contents - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a registering device for multimedia contents which automatically decides contents which are similar to contents having been registered so that they are not registered. SOLUTION: A database 2 provided on a server 1 comprises a contents database 20 which stores contents and a feature parameter database 21 which stores feature parameters of respective contents; when a portable terminal sends a registration request to the server 1, the server 1 extracts feature parameters of the contents requested to be registered and calculates their similarity to respective feature parameters in the feature parameter database 21. According to the calculated similarity, it is decided whether or not the contents are registered in the contents database 20.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus for registering various types of multimedia information (hereinafter, referred to as contents) such as sound, image, text, and the like.

[0002]

2. Description of the Related Art With the spread of the Internet and the improvement of the capabilities of terminals connected to the Internet, music distribution companies and image distribution companies have begun to distribute contents to each terminal for free or for a fee. I have. In addition, it has become possible to exchange contents between terminals or to register contents in a specific server, and to allow a user accessing the contents to download the contents for free or for a fee.

For example, in a system for distributing contents from a music distribution company to users, contents for each singer are registered on a Web site, and a user who accesses this site clicks contents of a desired singer. It can be downloaded for a fee or free of charge. In addition, the content created by the user himself / herself is registered in a specific Web site, and the user who accesses this site can download the content to his / her own terminal by clicking the title of the content for a fee or free of charge.

[0004]

However, in a system for distributing content from a music distribution company or an image distribution company according to a user's request, a royalty is usually paid to a copyright holder. Although it is considered that there is no problem under the copyright law, exchange of contents between general users and the like are not often treated under the copyright law.

[0005] However, if such a situation is left as it is, it is impossible to prevent the occurrence of troubles under the copyright law.
Further, if the registration of the content is prohibited, the distribution of the content is hindered, and the development of industry and culture via the network is hindered.

[0006] As a method for guaranteeing and authenticating the copyright of the content, a method using a digital watermarking technique and a method using an encryption technique have been proposed. In order to determine the similarity and eliminate the one that is very similar to the original content, a complicated algorithm for decoding the digital watermark or encryption is required, and the number of determination steps tends to be enormous.

[0007] Therefore, an object of the present invention is to automatically judge not to register contents having the same or similarity with already registered contents, to reduce the number of steps for the judgment, and to relatively reduce the number of steps. It is an object of the present invention to provide a multimedia content registration device that can easily perform a determination.

[0008]

Means for Solving the Problems The present invention is configured as follows in order to solve the above-mentioned problems. (1) A content database in which multimedia contents are registered, a characteristic parameter database in which characteristic parameters of each multimedia content registered in the content database are extracted and stored, and an attempt is made to newly register in the content database. Feature parameter similarity calculating means for extracting the feature parameters of the registered candidate multimedia content and calculating the similarity with each feature parameter in the feature parameter database; and calculating the similarity calculated by the feature parameter similarity calculating means. Registration determining means for determining, based on the registration candidate multimedia content, whether or not registration is appropriate;
It has.

According to the present invention, the server is provided with a feature database in which the feature parameters of the content are extracted and stored together with the content database in which the content is registered. From one user,
When the content to be registered (registration candidate content) is transmitted via the Internet or the like, the server extracts the characteristic parameters of the registration candidate content,
All similarities with each feature parameter stored in the feature parameter database are calculated. Of all the similarities obtained in this way, it is automatically determined whether there is a high similarity that is considered to infringe the copyright under the copyright law, and such a high similarity is determined. If so, the registration of the registration candidate content is rejected, and if not, it is automatically registered in the content database.

If the degree of similarity is neither high nor low and is in the gray zone (a range of similarity where it is not clear whether the infringement is infringed), the registration is suspended,
After warning the user, registration is performed, or registration permission or registration rejection is performed based on the judgment of the administrator.

(2) There are various methods for extracting feature parameters. In the simplest case, information for determining file size or file identity (a hash function such as MD5, a CRC (Cyclic Redundan)
cy Check)) as a feature parameter. For example, the file size of the original content or a hash value uniquely derived using a hash function such as MD5 is stored in a database, and by comparing these, the same content as the original content is registered. Can be prevented. Instead of using the hash value, the file identity can be almost determined by mutually comparing error detection codes such as CRC.
In the embodiment of the present invention, the accuracy of the identity determination is enhanced by further comparing the file size.

When the content is music, the motif of the music can be used as the feature parameter. The detection of the motif of a music is typically performed by extracting the melody data of the first four measures or the melody data from the first measure to the end of the next form, and extracting a short note or less than a specified value. The data string from which rests have been deleted can be used as the motif of the music. Furthermore, the lyric data of the song or the frequency of appearance of words based on the data may be used as the feature parameter.

[0013]

FIG. 1 is a configuration diagram of a content registration system according to an embodiment of the present invention.

The registration system comprises a server 1 provided with a database 2 in which contents and characteristic parameters are registered, and a terminal connected to the server 1. Although the terminal shows the portable terminals 1 and 2 in the figure, the present invention is not limited to this, and includes a terminal such as a personal computer. Each terminal is connected to the server 1.
Includes a database server, a Web server, and the like. The server and the terminal can be connected in a wired or wireless manner by a computer network such as the Internet or a public line.

In FIG. 1, a database 2 comprises a content database 20 and a feature parameter database 2
And 1. The content database 20 stores the content itself and the ID of the user who created the content.
Are stored in the feature parameter database 21.
The characteristic parameters extracted from the respective contents are stored in correspondence with the respective contents.

In FIG. 1, each portable terminal (portable terminal 1 and portable terminal 2) can create a content on each terminal. Now, when the portable terminal 1 creates a content and attempts to register it in the content database 20, it issues a registration request to the server 1. When receiving the registration request, the server 1 performs the following processing with the content as a registration candidate content.

(A) Extract feature parameters of registration candidate content.

(B) Each feature parameter stored in the feature parameter database 21 is compared with the feature parameter of the registration candidate content, and the similarity between them is calculated.

(C) Based on each similarity, whether or not the registration candidate content is registered is determined.

In (c) above, if there is no content having a high degree of similarity, registration permission is issued for the registration candidate content. As a result, the server 1 registers the content in the content database 20 and registers the characteristic parameter in the characteristic parameter database 21. Further, the portable terminal 1 is notified that the registration has been permitted.

Similarly, when the portable terminal 2 creates a content and issues a registration request to the server 1, the server 1 performs the above-described processes (a) to (c). If the similarity between the feature parameter of the registration candidate content and each feature parameter stored in the feature parameter database 21 has a high value, the registration of the content is rejected. That is, the content is not registered in the content database 20 and the portable terminal 2 is notified that the registration is rejected.

The server 1 automatically performs the above processing. As a result, there is no content database 20 having a high similarity to each other.

Next, the characteristic parameters extracted from the contents will be described. In this embodiment, it is assumed that all characteristic parameters extracted from the content relate to musical sound data.

FIG. 2 shows a first embodiment. FIG. 2 shows the configuration of the database 2. In this embodiment, a file size and a hash value of the content are extracted as characteristic parameters, and the extracted contents are used as the characteristic parameter database 2.
1 is stored. By making the feature parameters the file size and the hash value, the calculation of the similarity becomes extremely simple. That is, if the file size and the hash value are the same as the file size and the hash value of the registration candidate content, it can be estimated that both contents are exactly the same.

FIG. 3 shows an embodiment in which a file size for each part and a hash value for each part are extracted as feature parameters and are stored in the feature parameter database 21. In this embodiment, since the similarity for each part can be calculated, registration can be rejected even if any part such as a melody part or a base part is imitated.

Note that a CRC used for error code detection can be used instead of the hash value.

FIG. 4 shows an embodiment in which a motif is extracted as a feature parameter and stored in the feature parameter database 21. The motif of the music can usually be extracted from the melody part. There are various methods for extracting motifs.
The melody data of the bar can be extracted as the motif data, and the melody data from the first end bar to the first bar can be extracted as the motif data. Further, data obtained by deleting short notes or rests of a specified value or less from the data thus extracted can be used as motif data. In addition, discard the duration information,
It is also possible to use motif data only as a sequence of pitches and rests. It is also possible to disregard the order of appearance of pitches and rests, to summarize the frequency of appearance of each note, and use this as motif data. Further, it is also possible to delete short notes and rests of a specified value or less from this data and finally calculate the appearance frequency of each note as motif data. In addition, any method may be used as long as the method can represent the motif of the music by a data string.

FIG. 5 shows the structure of a database when motif data for each part is extracted as a feature parameter and stored in the feature parameter database 21. Similar to the embodiment of FIG. 3, since the motif data is stored for each part, it is possible to determine the presence or absence of imitation more finely.

FIG. 6 shows the configuration of a database in the case where data based on lyrics data is extracted as a feature parameter and stored in the feature parameter database 21.

As data based on the lyrics data, for example, a data string in which the lyrics data is all hiragana, or
Lyrics are divided for each word, and data obtained by summing up the appearance frequencies of the appearing words can be used as feature parameters. The embodiment shown in FIG. 6 is not used to calculate the similarity of a melody or the like of a song, but is used to calculate the similarity of lyrics.

A feature parameter database 21 as shown in FIGS. 2 to 6 is prepared in advance, and in this state, the feature parameters of the registration candidate contents requested to be registered by the terminal are stored in the feature parameter database 21. Each feature parameter is compared one by one. The calculation of the similarity is performed in the feature parameter database 21.
Is used as a reference parameter, the spatial distance of the registered candidate content feature parameter with respect to this reference parameter is calculated, and this value is output as a similarity. Various known methods such as a dynamic programming method for obtaining a distance and a hidden Markov method can be adopted. Preferably, the calculated similarities are segmented by using a predetermined threshold. For example, if the similarities are classified into A to C, A is a high similarity class, B is a medium similarity class, and C is a low similarity class. For example, as shown in FIG. A list of similarities to the content is displayed. In the drawing, No at the left end indicates a content number. In this example, the first and second contents have a low similarity (C), the third content has a high similarity (A), and the fourth content has an intermediate similarity. I have.
In addition, the fact that the similarity is intermediate (B) indicates that the image is in a gray zone where the similarity is neither high nor low.

In the server 1, if there is a similarity A in the list of similarities obtained in this way, the registration candidate content is not registered. In addition, similarity B
If there is, a holding process described later is performed. Only when all have the similarity C, the registration candidate content is registered.

When there is a content having a similarity level B, the registration candidate content is held. The suspension process in the present embodiment means a process of not registering the content immediately, but performing a registration after issuing a warning or notification to the user, or entrusting the decision to an administrator. How to perform the contents of the suspension processing depends on the calculated accuracy of the similarity, the operation rules of the system, and the like. For example, when the accuracy of the calculated similarity is low, it can be said that it is desirable to entrust the determination of the presence or absence of registration to the administrator in order to eliminate the risk of registering very similar contents. Conversely, when the accuracy is high, registration is performed, but it is considered that one method is to notify the user that the similarity is in the gray zone.

FIG. 8 is a flowchart showing the above processing in chronological order. This operation is performed in the server 1.

When there is a content registration request from a terminal (step ST1), the requested parameter is extracted as a registration candidate content and its characteristic parameters are extracted (ST2). Next, the similarity between the extracted feature parameters and all the feature parameters stored in the feature parameter database 21 is calculated, and a list as shown in FIG. 7 is created. Next, referring to the same list, it is determined whether there is the one having the similarity A (ST4) or the one having the similarity B (ST5). If there is one with similarity A, the registration is rejected, the terminal is notified to that effect, and the process ends (ST6). If there is no similarity A but a similarity B, a hold process is performed on the content (ST7), and the process ends. If none of the similarities A and B exist, the registration candidate content is registered in the content database 20, and the characteristic parameters are stored in the characteristic parameter database 21, and the registration is permitted to the terminal. A notification is made (ST8).

In the above operation, the feature parameters of the registration candidate content are extracted in the server 1, but this process may be performed in each terminal. In that case, when the terminal makes a registration request to the server 1, the content and the extracted characteristic parameters are transmitted simultaneously.

The content may be not only music information such as music, but also an image. In this case, the feature parameter of the image is stored in the feature parameter database, and the feature parameter of the image is extracted from the registration candidate content.

[0038]

According to the present invention, since the server is provided with the feature parameter database storing the feature parameters of each content together with the content database, the content requested to be registered by the terminal is stored in the content database. It is possible to easily determine whether or not the content is similar. For this reason, in particular, distribution including registration and use of contents via the Internet is promoted.

[Brief description of the drawings]

FIG. 1 is a schematic configuration diagram of a multimedia content registration system according to an embodiment of the present invention;

FIG. 2 is a configuration example of a database.

FIG. 3 shows another configuration example of a database.

FIG. 4 is another configuration example of a database.

FIG. 5 is another configuration example of a database.

FIG. 6 shows another configuration example of a database.

FIG. 7 shows a similarity list.

FIG. 8 is a flowchart showing a schematic operation of a server;

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat ゛ (Reference) G06F 17/30 350 G06F 17/30 350C 19/00 140 19/00 140

Claims (5)

[Claims] 1. A content database in which multimedia contents are registered, a feature parameter database in which feature parameters of each multimedia content registered in the content database are extracted and stored, and newly registered in the content database. A feature parameter similarity calculating means for extracting a feature parameter of the registration candidate multimedia content to be calculated and calculating a similarity with each feature parameter in the feature parameter database; and a similarity calculated by the feature parameter similarity calculating means. Registration determination means for determining whether the registration candidate multimedia content is suitable for registration based on the degree.
Multimedia content registration system. 2. The multimedia content registration system according to claim 1, wherein the characteristic parameter includes a file size or identity determination information of the multimedia content. 3. The multimedia content registration system according to claim 1, wherein said characteristic parameter is a motif of a music piece. 4. The multimedia content registration system according to claim 1, wherein the characteristic parameter is lyric data of music or an appearance frequency of a word based on the data. 5. The multimedia content registration system according to claim 1, wherein the registration suitability of the registration candidate multimedia content includes registration permission, registration rejection, and registration suspension.