JP2002041059A - Music content distribution system and method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a music content distribution system and method capable of reducing labor to search contents and creating opportunities to meet with unknown contents. SOLUTION: The music content distribution system 1 includes an off-line processing section 11 and an on-line processing section 12, and the off-line processing section 11 extracts acoustic parameters of the contents selected by a user and stores them to a user information database DB2. When the user accesses at the second time and later, the on-line processing section 12 extracts acoustic parameters of all music contents in a music contents database DB1 and exhibits recommendable music contents meeting the idea of the user by using the acoustic parameters stored in both the music contents database DB1 and the user information database DB2.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a music content distribution apparatus and method for distributing music content via a network.

[0002]

2. Description of the Related Art In recent years, with the standardization of data encoding systems, music content distribution services via computer networks such as the Internet have become widespread. With this distribution service, the latest contents that could only be obtained at a CD sales store or rental store can be quickly and easily obtained while staying at home or in the city.

[0003] Conventional music content distribution is mainly performed in the following two forms on a homepage on the Internet.

[0004] As a first distribution form, a music content provider opens a homepage, and collects and distributes a large amount of music content there. They are categorized and organized into genres (jazz, pop, rock, etc.) and are further categorized by artist in each genre. The user has searched for desired content and listened to the content to download the content based on the genre. It should be noted that a homepage that collects only music contents of a specific genre can be said to belong to this form. As a second distribution mode, an artist opens his / her own homepage and uploads and distributes created content. The user who likes the music of the artist accesses the homepage and downloads the favorite content. However, also in this case, in consideration of the user's operation of searching the Internet to find the homepage of one artist, it can be said that the music is classified and arranged by artist on the Internet.

[0005] Therefore, what can be said in common to these two distribution modes is that music contents are conventionally classified and arranged by genre or artist and presented to the user.

[0006] As described above, the user obtains music contents while staying at home or in the city, and appreciates music using a computer or a portable terminal.

[0007]

However, in today's world where the number of music contents uploaded on the Internet is increasing, the search operation for the desired music contents is extremely complicated for the user, and the labor becomes enormous. That is the current situation.

[0008] In addition, as music genres are diversified and integrated, it is pointed out that the classification of content providers is different from the classification of users, and that users cannot search for desired contents.

[0009] Further, even if the user performs a content search in accordance with the conventional classification, there are few opportunities to meet the content of a genre or an artist which has not been heard before, so that the musical taste of the user tends to be uniform. It is in.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned conventional problems, and has as its object to distribute music contents which can reduce the content search effort and create an opportunity to encounter unknown contents. An apparatus and a method thereof are provided.

[0011]

In order to solve the above-mentioned conventional problems, a music content distribution apparatus according to a first aspect of the present invention transmits a music content selected by a user from stored music contents to a network. A music content distribution device for distributing the selected music content as a first sound parameter, and user information storing at least the first sound parameter. Storage means, second sound parameter extracting means for extracting sound parameters of the stored music contents as second sound parameters when the user selects music contents again, and re-selection of the music contents by the user. Upon selection, the first acoustic parameter stored in the user information storage means is selected. And a recommended content search / presentation means for searching for and presenting recommended music contents from the stored music contents using the extracted second audio parameters. .

According to the first aspect of the present invention, when the first acoustic parameter extracting means extracts the acoustic parameter of the music content selected by the user as the first acoustic parameter, the first acoustic parameter is It is stored in the user information storage means, and when the user accesses the apparatus and selects and downloads music content again, the second sound parameter extraction means
The acoustic parameter of the stored music content is extracted as the second acoustic parameter, and the recommended content search / presentation unit divides the first acoustic parameter stored in the user information storage unit and the extracted second acoustic parameter. In addition, recommended music contents are retrieved from the stored music contents and presented to the user, so that music contents that match the user's tastes can be recommended, and the user is able to eliminate the complexity of the search operation. You.

A music content distribution device according to a second aspect of the present invention is the music content distribution device according to the first aspect, wherein the recommended content search / presentation means includes a statistic using sound parameters as explanatory variables when searching for the music content. It is characterized by using a dynamic method.

According to the second aspect of the present invention, it is possible to improve the search accuracy of music content that suits the user's preference by using a statistical method using acoustic parameters as explanatory variables.

[0015] A music content distribution device according to a third aspect of the present invention is the device according to the first or second aspect,
A third acoustic parameter extracting unit configured to extract, as a third acoustic parameter, an acoustic parameter of the music content which has been presented to the user but has not been selected; and a third acoustic parameter extracting unit configured to extract the extracted third acoustic parameter and the first acoustic parameter. Weighting conversion coefficient calculating means for calculating a weighting conversion coefficient in advance using the acoustic parameters of the, the recommended content search and presentation means, when searching for the music content,
The second acoustic parameter is multiplied by the weighting conversion coefficient.

According to the third aspect of the present invention, the third acoustic parameter extracting means extracts, as the third acoustic parameter, an acoustic parameter of the music content which has been presented to the user but has not been selected. On the other hand, the weighting conversion coefficient calculating means calculates the weighting conversion coefficient in advance using the third sound parameter and the first sound parameter, and when the user selects the content again, Recommended content search presentation means,
In searching for music content, the second acoustic parameter is multiplied by a weighting conversion coefficient, so that the user's taste can be more strongly reflected in the search result and presented.

A music content distribution device according to a fourth aspect of the present invention is the music content distribution device according to the third aspect, wherein the weighting transform coefficient is a variance within the acoustic parameter when there are a plurality of the first acoustic parameters. When the plurality of third acoustic parameters are present, the ratio with the variance in the acoustic parameters is maximized.

According to the fourth aspect of the present invention, the user's preferences can be more accurately reflected in the search results.

A music content distribution apparatus according to a fifth aspect of the present invention is the apparatus according to any one of the first to fourth aspects, wherein the extracted acoustic parameters are cepstrum, audio power, sounding time vector. Or at least one of the following.

According to the fifth aspect of the present invention, by using the above parameters, it is possible to present a content reflecting the user's taste.

According to a sixth aspect of the present invention, in the music content distribution method according to the first aspect of the present invention, a music content selected by a user from among stored music contents is distributed via a network. Extracting and storing the acoustic parameter of the music content obtained as the first acoustic parameter, and extracting the acoustic parameter of the stored music content as a second acoustic parameter when the user selects the music content again. And when the user selects the music content again, using the first acoustic parameter stored in the user information storage means and the extracted second acoustic parameter, the stored music. Searching and presenting recommended music content from the content; Characterized by comprising.

[0022]

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, embodiments of a music content distribution apparatus and method according to the present invention will be described in detail with reference to the drawings.

First, the outline of the embodiment will be described. In the apparatus according to the present embodiment, different processes are performed for the first user access and the second and subsequent user accesses. In the first time, all music contents held by a music content provider or the like (hereinafter abbreviated as provider as appropriate) are presented to the user, and acoustic parameters are obtained from both the content obtained (downloaded) by the user and the content not obtained. The user's taste is learned by obtaining a parameter conversion coefficient that maximizes the F ratio offline.

In the second and subsequent accesses, all music contents are discriminated and analyzed by using the acoustic parameters of the contents which the user has learned up to the previous time and the contents of the contents which the user does not like and the above-mentioned parameter conversion coefficient. And dislike them, and present the former to the user.

The user selects and obtains a desired content from the contents. It should be noted that the content acquired by the user at this time is also learned by the offline processing so that the learning content can be reflected in the subsequent processing.

With the above-described processing, the apparatus of the present embodiment can present music contents that match the user's preferences regardless of the genre.

FIG. 1 is a diagram showing a configuration and a use form of a content distribution server 1 to which an embodiment of the present invention is applied. Content distribution server 1 operated by provider
Is a computer connected to the Internet NET via a gateway or a router (both not shown). On the other hand, Internet service providers (IS
A user terminal 3 having a web browser is connected to the computer 2 of P). Content distribution server 1
Is capable of transmitting and receiving data to and from the user terminal 3 using the TCP / IP protocol via the Internet NET. Other user terminals connected to the Internet NET are not shown.

The content distribution server 1 has, on a connected storage device, a music content database DB1 in which all contents held by the provider are stored and a user information database DB2 in which user information is stored.

The content distribution server 1 presents content that suits the user's taste from all the music contents in the music content database DB1, but in consideration of the user's convenience, the content distribution server 1 may also use an old genre. All music content can be classified as genre I
D is associated therewith. User information database DB
2, the member number, name, address, e-mail address, and the like of the member (user) who has contracted with the content provider are registered in advance.

The content distribution server 1 includes an offline processing unit 11 for performing processing when a user is offline and an online processing unit 12 for performing processing when online.
It has. These processing units are realized by a CPU executing a control program stored in a storage device, and a part of the functions is implemented by a DSP (Digital Signal Processor).
(sor) may be used.

The offline processing unit 11 includes an audio parameter extracting unit 111 for extracting audio parameters for each of the music content selected by the user in the past and the music content presented but not selected despite being presented. And a parameter conversion coefficient calculation unit 112 (weight conversion coefficient calculation means) for calculating a parameter conversion coefficient which is a weighted conversion coefficient. The extracted acoustic parameters and the calculated parameter conversion coefficients are stored in the user information database DB2 for each user. Further, the acoustic parameter extracting unit 111 has a configuration combining the first acoustic parameter extracting unit and the third acoustic parameter extracting unit of the present invention.

The online processing section 12 extracts an acoustic parameter from all contents held by the provider, and a parameter for converting the extracted acoustic parameter by a parameter conversion coefficient stored in the user information database DB2. Conversion unit 122
And a discriminant analysis unit 1 that performs discriminant analysis of music content using a statistical method using the converted parameters as explanatory variables.
23. The sound parameter extraction unit 121
This corresponds to a second acoustic parameter extracting unit of the present invention.

FIG. 2 is a flowchart showing processing for the first user access.

In step S1, when a search request for music content is made by the user, all music contents in the music content database DB1 are presented on the screen of the user terminal 3 in step S3. This presentation and similar content presentation described below can be easily performed by transmitting an HTML document using the http protocol.

In step S5, when the user selects a desired content from the presented contents, downloading of the selected content is started. When the download is completed, in step S7, the offline processing unit 1
1 performs offline processing. In the offline processing, learning of acoustic characteristics of music content is performed as described below.

FIG. 3 is a flowchart showing details of the offline processing in step S7. Offline processing unit 1
The first acoustic parameter extraction unit 111 performs step S11.
Then, sound parameters (first sound parameters) are obtained from each of the contents selected by the user, and stored in the user information database DB2 in association with the contents. On the other hand, in step S13, the sound parameters (third sound parameters) of the respective contents which are not selected despite being presented are obtained and stored in the user information database DB2 in association with the contents. In addition,
The reason why the acoustic parameters are not extracted only for the downloaded (acquired) content but the content selected by the user is to be extracted is because download failure is considered. That is, even if there is a download failure, the parameter extraction of the content selected by the user is not interrupted / stopped.

Here, the sound parameter extraction processing in the sound parameter extraction unit 111 will be described in more detail.

FIG. 4 is a flowchart showing a process of extracting sound parameters from one music content. Note that the same processing is performed in the acoustic parameter extraction unit 121 as well.

First, in step S21, audio power is determined frame by frame from music content (frame length: about 30 ms, frame shift: about 5 ms is preferable). Next, in step S23, a section having a high audio power (a section of about 10 seconds) is extracted. This is regarded as one section of the rust (rust part). The audio power of the chorus part is an element of the acoustic parameter.

Then, in step S25, the music content is band-divided by the filter bank. The number of divisions is preferably about 10 to 20. Then, step S2
In 7, the cepstrum is extracted for each output of each band.

The order of the cepstrum is preferably set to about 30 when the sampling frequency of the music data is 44.1 kHz. Although not shown, a delta (differential) cepstrum is also extracted.

In parallel with the processing in step S27, in step S29, a sounding time vector is extracted from the filter output for each band, and in step S31, the sounding time vector intensity is calculated from the extracted vector.

The algorithm for extracting the onset time vector is as follows:
For example, (Yoichi Nagashima, Shuji Hashimoto, Joe Hiraga, Keiji Hirata,
"The World of Computers and Music: From Fundamentals to Frontiers", Kyoritsu Shuppan, ISBN4-320-02942-9
(August 1999)) (Goto, “Beat Recognition (Beat Tracking)”, Chapter 3, Section 3, pp. 100-116) can be used.

FIG. 5 is a diagram showing a configuration of acoustic parameters extracted from contents.

Regarding the cepstrum of each band, one vector (the number of elements: the same as the cepstrum order) is obtained for each content by averaging in the time direction, and is used as an element of the acoustic parameter. Similarly, one vector (the number of elements: the same as the cepstrum order) is obtained for each content by averaging in the time direction, and the delta cepstrum is used as an element of the acoustic parameter.

Regarding the sound power, the first to fourth moments are obtained and used as acoustic parameters. The first moment corresponds to the average value, the second moment corresponds to the variance, the third moment corresponds to the skewness, and the fourth moment corresponds to the kurtosis.

Similarly, the first to fourth moments of the delta (difference) of the audio power are calculated. Finally, the first to fourth moments are calculated for the onset time vector intensity. As described above, the order of the extracted acoustic parameters is 2XNXC + 12. Where N is the number of band divisions,
C represents the cepstrum order.

Returning to FIG. 3, the description will be continued. Step S1
In step 5, a parameter conversion coefficient λ is calculated using the extracted acoustic parameters, and stored in the user information database DB2 in association with the content. Here, in the calculation of the transform coefficient, the F ratio, which is the ratio between the intra-group variance and the out-group variance, is set to be maximum. The F ratio is a statistical parameter represented by the following equation (1).

[0049]

(Equation 1) In the equation (1), m represents the number of groups to be determined. As in the present embodiment, when there are two groups of the music content selected by the user and a group other than the group, m = 2. I do.

Xik is the sound parameter of the k-th content in the content group selected by the user (when i = 0) or the content group not selected (when i = 1), and μi is the user. The average value of the acoustic parameters of the selected content group (if i = 0) or the unselected content group (if i = 1)
μ represents an average value of acoustic parameters of all presented music contents.

The larger the value of the F ratio, the more effective the parameter for discriminating the contents.

More specifically, in order to generate a valid parameter set with a small number by linearly combining a large number of parameters using the F ratio as a reference, an eigen equation such as the following equation (2) is solved. To obtain the parameter conversion coefficient λ.

[0053]

(Equation 2) In Expression (2), B represents a covariance matrix between a content group selected by the user and a non-selected content group, and W represents a covariance matrix in both content groups. The definitions are shown in the following equations (3) and (4).

[0054]

(Equation 3) Note that the obtained parameter conversion coefficient λ is a matrix of the number of acoustic parameters × the number of acoustic parameters.

FIG. 6 is a flowchart showing the flow of processing for the second and subsequent user accesses. In step S51, when there is a music content search request from the user, in step S53, the online processing unit 12 performs online processing. In the online processing, as described below, it is determined whether or not all music contents in the music content database DB1 match the user's taste. FIG. 7 is a flowchart showing the offline processing of step S53.

In step S61, audio parameters (second audio parameters) of each content in the music content database DB1 are extracted. Note that such processing can be performed even when offline. In the following step S63, the acoustic parameters extracted in step S61 are converted using the user's parameter conversion coefficient λ stored in the user information database DB2. Specifically, a matrix calculation of multiplying (multiplying) the parameter conversion coefficient λ by the acoustic parameter is performed. By performing this processing for all contents in the music content database DB1, sound parameters for the number of contents are obtained.

Next, in step S65, the discriminant analysis unit 1
23 performs discriminant analysis using the Mahalanobis distance, using each of the parameters for the number of contents as an explanatory variable, and from among all contents, the closest to the center of gravity of the group of contents selected by the user in the past on the Mahalanobis distance scale For example, about 100 contents are extracted in order.
The music content extracted here is the music content that the user would like. It should be noted that subsequent processing is not performed for music contents that the user does not like.

In step S67, the contents extracted in step S65 are converted based on the corresponding genre ID.
Categorized into traditional music genres such as jazz and rock.

Returning to FIG. 6, in step S55, the online processing section 12 presents the classified contents to the user. Steps S53 and S55 correspond to the recommended content search / presentation means of the present invention.

FIG. 8 is a diagram showing a configuration example of a screen for presenting music content to a user. In FIG. 8, on the left side of the screen, music contents that the user is likely to like are arranged and presented according to old genres on the left side of the screen as “automatic detection results”. The user can select and acquire a desired content from the contents by operating a mouse or the like. On the other hand, on the right side of the screen, a screen for inputting a keyword such as a content name or an artist name is provided as a “general search screen” so that music content that has not been presented can be searched. Further, as in the lower right of the screen, a new content recently stored in the music content database DB1 may be presented.

Returning to FIG. 6, when the user selects a desired content from the presented contents in step S57, the download of the selected content is started. Subsequently, in step S59, the offline processing (learning) shown in FIG. 3 is performed, and the user information database D is obtained using the obtained acoustic parameters and parameter conversion coefficients.
The user information in B2 is updated.

In the present embodiment, the searched contents are further classified according to the old music genre in consideration of the convenience of the user. However, if such additional processing is omitted, the genre ID and the like can be obtained. Attribute information is unnecessary,
Music content can be handled centrally.

It should be noted that the online processing section 12 can also be configured to search the music content database DB1 for a music content whose acoustic parameters are similar to the music content selected by the user in the past, and present it to the user. The processing using the conversion coefficient λ can be omitted.

In the first user access, all the music contents in the music content database DB1 are presented. However, for example, several songs may be selected and presented from the old genre.

The position of the content distribution server 1 on the Internet is not limited at all, and may be provided on a private network (LAN) as needed. At that time, protocols other than TCP / IP can be used.

As described above, according to the music content distribution apparatus or method according to the embodiment of the present invention, the sound parameters of the music content selected by the user are extracted and stored, and the music content by the user is stored. When the user selects again, using the acoustic parameters extracted from the stored music contents and the stored acoustic parameters, the system is configured to search for and present recommended music contents. Becomes possible.

As a result, the user can select music content regardless of the old genre, thereby reducing the content search effort and increasing the chances of encountering content of a different genre. Can be expanded.

[0068]

As described above, according to the music content supply apparatus or the music content supply method of the present invention, the sound parameters of the music content selected by the user are extracted and stored, and the music content is again reproduced by the user. When selecting a music content, the recommended music content is searched and presented using the sound parameters extracted from the stored music contents and the stored sound parameters, so that the content search effort can be reduced and the opportunity to encounter unknown contents can be reduced. Creation becomes possible.

[Brief description of the drawings]

FIG. 1 is a diagram showing a configuration and a use form of a content distribution server 1 to which an embodiment of the present invention is applied.

FIG. 2 is a flowchart illustrating processing for a first user access.

FIG. 3 is a flowchart showing off-line processing.

FIG. 4 is a flowchart showing an acoustic parameter extraction process.

FIG. 5 is a diagram showing a configuration of acoustic parameters.

FIG. 6 is a flowchart of a process for the second and subsequent user accesses.

FIG. 7 is a flowchart showing off-line processing.

FIG. 8 is a diagram showing a configuration example of a music content presentation screen.

[Explanation of symbols]

1 Content Distribution Server (Music Content Distribution Device) 2 Computer of Internet Service Provider 3 User Terminal 11 Offline Processing Unit 12 Online Processing Unit 111 Sound Parameter Extraction Unit (First Sound Parameter Extraction Means, Third Sound Parameter Extraction Means) 112 Parameter conversion coefficient calculation unit 121 Sound parameter extraction unit (second sound parameter extraction unit) 122 Parameter conversion unit 123 Discriminant analysis unit DB1 Music content database DB2 User information database NET Internet

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat ゛ (Reference) G10L 15/02 G10L 3/00 531N // G10L 101: 04 9/16 301A F-term (Reference) 5B075 KK03 KK07 KK13 KK33 KK37 ND03 ND14 ND23 ND36 PQ04 PQ05 PQ75 PR03 PR08 QP10 UU40 5D015 FF04 HH05

Claims (6)

[Claims] 1. A music content distribution apparatus for distributing, via a network, a music content selected by a user from among stored music contents, wherein a sound parameter of the selected music content is set to a first parameter.
First acoustic parameter extracting means for extracting as the acoustic parameters, user information storage means for storing at least the first acoustic parameters, and when the user selects music content again,
The acoustic parameter of the stored music content is
A second acoustic parameter extracting means for extracting as the acoustic parameters of: when the user selects music content again,
A recommendation for searching for and presenting recommended music contents from the stored music contents using the first sound parameters stored in the user information storage means and the extracted second sound parameters. A music content distribution device comprising: a content search / presentation unit. 2. The music content distribution apparatus according to claim 1, wherein the recommended content search / presentation unit uses a statistical method using acoustic parameters as explanatory variables when searching for the music content. 3. An acoustic parameter of a music content which has not been selected despite being presented to the user is stored in a third area.
A third acoustic parameter extracting means for extracting the acoustic parameters as the acoustic parameters; and a weighting transform coefficient calculating means for calculating a weighting transform coefficient in advance using the extracted third acoustic parameters and the first acoustic parameters. 3. The music content distribution device according to claim 1, wherein the recommended content search / presentation unit multiplies the second acoustic parameter by the weighting conversion coefficient when searching for the music content. 4. The weighted transform coefficient sets a maximum ratio between the variance in the acoustic parameter when there are a plurality of the first acoustic parameters and the variance in the acoustic parameter when there are a plurality of third acoustic parameters. The music content distribution device according to claim 3, wherein 5. The music content distribution apparatus according to claim 1, wherein the extracted acoustic parameters include at least one of cepstrum, audio power, and onset time vector. 6. A method for distributing music content selected by a user from among stored music contents via a network, wherein the audio parameter of the selected music content is set to a first parameter.
Extracting and storing as acoustic parameters of the music content; and when the user selects music content again,
The acoustic parameter of the stored music content is
Extracting the audio content as a sound parameter, and when the user selects music content again,
A step of searching for and presenting recommended music contents from the stored music contents using the first sound parameters stored in the user information storage means and the extracted second sound parameters. A method for distributing music content, comprising: