JP2010165234A - Musical piece visualization device, musical piece visualization method, and musical piece visualization program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To allow a user to grasp a musical piece data distribution intuitively, facilitate the user's understanding of musical pieces stored in a database, and reduce the time required for musical piece retrieval. <P>SOLUTION: A musical piece visualization device 10 that displays the features of musical piece data visually comprises a feature amount extraction part 12 for reading musical piece data from a database 11 and extracting a plurality of kinds of feature amounts from the read musical piece data, a feature amount analysis part 13 for analyzing the distribution degree when each of the extracted feature amounts is shown on a one-dimensional coordinate axis, a coordinate axis selection part 14 for selecting one or more coordinate axes based on the distribution degree, and a display part 15 for displaying the distribution of the musical piece data using the selected coordinate axis and the feature mount corresponding to the coordinate axis. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a technique for visually displaying a distribution of music data corresponding to a feature value based on a feature value of music data.

  Conventionally, a technique for visualizing information is known. For example, in Patent Document 1, a singular vector is generated by performing singular value decomposition on a feature value digitized in a database. Then, the dimensional reduction is performed using the singular vector of the feature quantity, and a predetermined number of elements of the selection vector reduced in dimension are selected in descending order from the larger one to obtain a visualization coordinate axis.

  Moreover, in patent document 2, the visualization of a content is implemented by using the multidimensional analysis result obtained from the content using one color display. Specifically, an analysis result having a three-dimensional feature amount is mapped to a color space defined by RGB values.

  Further, in Patent Document 3, an impression amount based on a human sense is extracted from an acoustic feature amount such as a tempo of the music and an average number of sounds, and the impression information (sad, intense, etc.) is used as a coordinate axis to extract the music. Expressed on a dimensional map.

  Moreover, the following is mentioned as a music search system which applied the data visualization method. Non-Patent Document 1 discloses a technique for mapping music data two-dimensionally using a self-organizing map. Non-Patent Document 2 discloses a technique for extracting multidimensional acoustic features from music and mapping the music in two dimensions or three dimensions using a multivariate analysis method such as principal component analysis. . As tools for extracting feature quantities that can be used for visualization, there are tools as shown in Non-Patent Documents 3 and 4.

JP 2003-141165 A JP 2004-295293 A JP 2005-31644 A

E.Pampalk, S.Dixon, and G. Widmer, “Exploring music collections by browsing different views,” in ISMIR2003, pp.201-208 Hamawaki et al. "Music recommendation system using visualization technology", IPSJ Spring National Convention, 3N-9 jAudio: http://sourceforge.net/projects/jaudio/ MARSYAS: http://marsyas.sness.net/

  However, in the techniques described in Patent Documents 1 and 2, multidimensional feature values (vectors) obtained by analyzing content can be mapped in two dimensions and three dimensions. It is not easy to catch. For this reason, it is difficult to grasp the intuitive content distribution unless the user is familiar with the method or has not been able to understand the map by using the system for a while.

  On the other hand, in Patent Document 3, it is possible to solve the above problem by using an amount of impression based on human sense as a coordinate axis. However, when there is a bias in the characteristics of the data to be mapped (data in the database), there is a bias in a certain area of the map, and there is a problem that it is difficult to catch a difference in a minute area of the data group. . For example, if all the data is “bright” and “violent” data on the two axes “bright-dark” and “violent-admissive”, it is difficult to further narrow down the information. It's hard to say that it's kind.

  In Non-Patent Documents 1 and 2, multi-dimensional data extracted from music can be mapped to two-dimensional or three-dimensional using a self-organizing map or multivariate analysis. The axis itself has no meaning. In multivariate analysis, the meaning of the axis must be inferred from many data distributions in order to grasp the meaning of the axis itself. For this reason, in order for a user to search for the music which he wants to ask for, he has to understand a map, and there is a problem that it takes time to search.

  The present invention has been made in view of such circumstances, and allows the user to intuitively grasp the distribution of music data, promote understanding of music stored in the database, and search for music. An object of the present invention is to provide a music visualization device, a music visualization method, and a music visualization program that can reduce the time required for the music.

  (1) In order to achieve the above object, the present invention takes the following measures. That is, the music visualization apparatus of the present invention is a music visualization apparatus that visually displays the characteristics of music data, and reads music data from a database, and extracts a plurality of types of feature values from the read music data. A quantity extraction unit, a feature quantity analysis unit for analyzing the degree of distribution when each extracted feature quantity is represented on a one-dimensional coordinate axis, and one or more coordinate axes are selected based on the degree of distribution And a display unit that displays the distribution of music data using the selected coordinate axis and the feature amount corresponding to the coordinate axis.

  In this manner, the degree of distribution when the feature values extracted from the music data are each represented on a one-dimensional coordinate axis is analyzed, and one or more coordinate axes are selected based on the degree of distribution, and the selected coordinate axis And the distribution of music data is displayed using the feature quantity corresponding to the coordinate axis, so the coordinate axis itself can be given meaning, and even if the music data stored in the database is biased, the meaning of the coordinate axis It is possible to select the optimum coordinate axis for the target database. In addition, the user can intuitively understand the distribution of music data. In addition, it is possible to shorten the time required for the user's music search and reduce the burden, and to promote understanding of the music stored in the database.

  (2) Further, in the music visualization device of the present invention, the music visualization device further includes a classification processing unit that classifies the music data stored in the database or each feature amount extracted by the feature amount extraction unit, and the display unit includes: The classification result is displayed.

  As described above, since the music data stored in the database or each feature quantity extracted by the feature quantity extraction unit is classified, efficient music search can be performed even if the data quantity is enormous. .

  (3) Moreover, in the music visualization apparatus of the present invention, the coordinate axis selection unit selects any of a plurality of coordinate axes as coordinate axis candidates based on the distribution degree, and the display unit selects a user operation as a trigger. The distribution of music data is displayed using the coordinate axes and the feature amounts corresponding to the coordinate axes.

  Thus, based on the degree of distribution, any one of a plurality of coordinate axes is selected as a coordinate axis candidate, and the distribution of the music data is displayed using the selected coordinate axis and the feature amount corresponding to the coordinate axis triggered by the user's operation. Therefore, the user can select the coordinate axis according to his / her preference, and the distribution of the music data can be displayed with the selected coordinate axis and the feature amount corresponding to the coordinate axis. Thereby, it becomes possible to reduce the burden of the user in music search.

  (4) Further, the music visualization method of the present invention is a music visualization method for visually displaying the characteristics of music data, and in the feature amount extraction unit, the music data is read from the database, and from the read music data, A step of extracting a plurality of types of feature quantities; a step of analyzing the degree of distribution when each of the extracted feature quantities is represented on a one-dimensional coordinate axis in the feature quantity analysis section; and the distribution in the coordinate axis selection section Selecting one or more coordinate axes based on the degree; and displaying a distribution of music data using the selected coordinate axes and feature quantities corresponding to the coordinate axes in the display unit; It is characterized by including at least.

  In this manner, the degree of distribution when the feature values extracted from the music data are each represented on a one-dimensional coordinate axis is analyzed, and one or more coordinate axes are selected based on the degree of distribution, and the selected coordinate axis And the distribution of music data is displayed using the feature quantity corresponding to the coordinate axis, so the coordinate axis itself can be given meaning, and even if the music data stored in the database is biased, the meaning of the coordinate axis It is possible to select the optimum coordinate axis for the target database. In addition, the user can intuitively understand the distribution of music data. In addition, it is possible to shorten the time required for the user's music search and reduce the burden, and to promote understanding of the music stored in the database.

  (5) In the music visualization method of the present invention, the classification processing unit classifies the music data stored in the database or each feature amount extracted by the feature amount extraction unit, and the display unit And a step of displaying the classification result.

  As described above, since the music data stored in the database or each feature quantity extracted by the feature quantity extraction unit is classified, efficient music search can be performed even if the data quantity is enormous. .

  (6) The music visualization program according to the present invention is a music visualization program for visually displaying the characteristics of music data. The music data is read from a database, and a plurality of types of feature values are obtained from the read music data. A process of extracting, a process of analyzing a distribution degree when each of the extracted feature quantities is represented on a one-dimensional coordinate axis, and a process of selecting any one or more coordinate axes based on the distribution degree; A series of processes including a process of displaying a distribution of music data using the selected coordinate axis and a feature amount corresponding to the coordinate axis, and a computer-readable and executable command. It is said.

  In this way, the degree of distribution when the feature quantities extracted from the music data are each represented on a one-dimensional coordinate axis is analyzed, and one or more coordinate axes are selected and selected based on the degree of distribution. And the distribution of music data is displayed using the feature quantity corresponding to the coordinate axis, so the coordinate axis itself can be given meaning, and even if the music data stored in the database is biased, the meaning of the coordinate axis It is possible to select the optimum coordinate axis for the target database. In addition, the user can intuitively understand the distribution of music data. In addition, it is possible to shorten the time required for the user's music search and reduce the burden, and to promote understanding of the music stored in the database.

  (7) Further, the music visualization program of the present invention further includes a process of classifying the music data stored in the database or each extracted feature amount, and a process of displaying the classification result. It is a feature.

  As described above, since the music data stored in the database or each feature quantity extracted by the feature quantity extraction unit is classified, efficient music search can be performed even if the data quantity is enormous. .

  According to the present invention, since the coordinate axis itself is given meaning and the optimum axis is selected for the target database, it is possible to intuitively understand the data distribution. In addition, by applying the method to a search system, it is possible to reduce the time required for the user to search for music and promote music understanding in the database.

It is a block diagram which shows schematic structure of the music visualization apparatus which concerns on 1st Embodiment. It is a flowchart which shows operation | movement of the music visualization apparatus which concerns on 1st Embodiment. It is a block diagram which shows schematic structure of the music visualization apparatus which concerns on 2nd Embodiment. It is a flowchart which shows operation | movement of the music visualization apparatus which concerns on 2nd Embodiment. It is a figure which shows the outline of a music search system. It is a figure which shows an example of a user interface.

  Embodiments of the present invention will be described below with reference to the drawings. In the present embodiment, when two-dimensional or three-dimensional mapping of multidimensional data is performed, a plurality of coordinate axes that are easy to understand with human sense are selected, and the coordinate axes are adaptively changed according to the data bias. As a result, an intuitive understanding of the map can be promoted, and a data visualization technique that does not depend on data bias is provided. A user-friendly music search system is also provided.

(First embodiment)
FIG. 1 is a block diagram illustrating a schematic configuration of a music visualization device according to the first embodiment, and FIG. 2 is a flowchart illustrating an operation of the music visualization device according to the first embodiment. As shown in FIG. 1, the music visualization device 10 includes a database 11, a feature amount extraction unit 12, a feature amount analysis unit 13, a coordinate axis selection unit 14, and a display unit 15. The database 11 stores a plurality of music data. The feature amount extraction unit 12 extracts feature amounts from the music data (group) stored in the database 11 (step S10). For example, rhythm, tempo, gender of vocals, timbre, tonality, etc. can be cited as classification of feature quantities that are easy to understand intuitively by humans. This feature amount can be extracted using the technique disclosed in Non-Patent Document 3 or 4. The classification names listed here are called coordinate axis candidates.

  The feature quantity analysis unit 13 analyzes the degree of distribution on the coordinate axes with respect to the feature quantities that are coordinate axis candidates extracted by the feature quantity extraction unit 12 (step S11). For example, dimensional compression such as multivariate analysis can be applied to each coordinate axis candidate, and the coordinate degree can be analyzed by one-dimensional compression. Hereinafter, the score of the degree of distribution is X. If X is large, the bias is small, and if X is small, the bias is large. The technique disclosed in Patent Document 2 is originally based on a three-dimensional feature value, but in this embodiment, coordinate axis parameters are converted to one-dimensional from feature values having different dimensionality and characteristics. Thereby, it is possible to adapt to various feature-values.

  The coordinate axis selection unit 14 selects an axis from the coordinate axis candidates based on the degree of distribution of the coordinate axis candidates obtained by the feature amount analysis in the feature amount analysis unit 13 (step S12). For example, in the case of two-dimensional display, two coordinate axes that are higher in the score X can be selected, and in the case of three-dimensional display, three coordinate axes can be selected. In addition, by normalizing each coordinate axis, it is possible to cope with a case where a deviation occurs in an axis with a higher score. Thereby, even if the music data stored in the database 11 is biased, it is possible to select the optimal coordinate axis for the database 11 to be searched while retaining the meaning of the coordinate axis.

  The display unit 15 displays the music data based on the coordinate axis selected by the coordinate axis selection unit 14 (step S13). This makes it possible to use a one-dimensional parameter by dimension compression of multi-dimensional feature quantities classified as easy to understand intuitively as a coordinate axis, and to select an optimal axis for music data in the database 11 for further display. Therefore, it is possible to realize the visualization of music data that does not depend on the target database and is easy to understand intuitively.

(Second Embodiment)
FIG. 3 is a block diagram showing a schematic configuration of the music visualization device according to the second embodiment, and FIG. 4 is a flowchart showing the operation of the music visualization device according to the second embodiment. When the amount of music data stored in the database 11 is enormous and the number of display data increases on the map, it is displayed in two stages. For this reason, in the second embodiment, a classification processing unit 16 is added to the first embodiment shown in FIG.

  The classification processing unit 16 performs category classification for the music in the entire database 11 in advance. For example, using the feature quantity group obtained by the feature quantity extraction step (step S20), clustering is applied to classify into categories. In addition, when meta information is obtained in advance, the data in the database 11 can be classified into categories in advance based on the meta information. This category may be a rough category.

  By applying a feature amount analysis step (step S23), a coordinate axis selection step (step S24), and a secondary display step (step S25) for each category classified by such a classification process, a huge amount of data is applied. Also, the same merit as the first embodiment can be obtained.

  Note that the classification processing unit 16 can perform the classification process first without extracting the feature amount when it is not necessary to extract the feature amount, for example, when using meta information. That is, the order of processing can be changed as appropriate.

  Next, the operation of the music visualization device according to the second embodiment configured as described above will be described. As shown in FIG. 4, the feature quantity extraction unit 12 extracts feature quantities from the music data (group) stored in the database 11 (step S20). For example, rhythm, tempo, gender of vocals, timbre, tonality, etc. can be cited as classification of feature quantities that are easy to understand intuitively by humans. This feature amount can be extracted using the technique disclosed in Non-Patent Document 3 or 4. The classification names listed here are called coordinate axis candidates.

  The classification processing unit 16 performs category classification in advance on the music in the entire database 11 (step S21). For example, using the feature quantity group obtained by the feature quantity extraction step (step S20), clustering is applied to classify into categories.

  Next, in the second embodiment, primary display is performed on the display unit 15 (step S22). This primary display displays the category classification obtained by the classification processing step (step S21). For example, when clustering is applied using a feature quantity group, it can be realized by using multivariate analysis, a self-organizing map, or the like, which is an existing technique. When classification based on meta information is used, it can be expressed in two dimensions based on a list display or meta information. Thereby, it becomes possible to grasp a rough classification of a huge music data group, and data in the secondary display step (step S25) can be limited. A plurality of categories to be displayed in the secondary display step can be designated.

  Next, the feature quantity analysis unit 13 analyzes the degree of distribution on the coordinate axes with respect to the feature quantities that are coordinate axis candidates extracted by the feature quantity extraction unit 12 or classified by the classification processing unit 16 (step S23). ). For example, dimensional compression such as multivariate analysis can be applied to each coordinate axis candidate, and the coordinate degree can be analyzed by one-dimensional compression. Hereinafter, the score of the degree of distribution is X. If X is large, the bias is small, and if X is small, the bias is large. The technique disclosed in Patent Document 2 is originally based on a three-dimensional feature value, but in this embodiment, coordinate axis parameters are converted to one-dimensional from feature values having different dimensionality and characteristics. Thereby, it is possible to adapt to various feature-values.

  The coordinate axis selection unit 14 selects an axis from the coordinate axis candidates based on the degree of distribution of the coordinate axis candidates obtained by the feature amount analysis in the feature amount analysis unit 13 (step S24). For example, in the case of two-dimensional display, two coordinate axes that are higher in the score X can be selected, and in the case of three-dimensional display, three coordinate axes can be selected. In addition, by normalizing each coordinate axis, it is possible to cope with a case where a deviation occurs in an axis with a higher score. Thereby, even if the music data stored in the database 11 is biased, it is possible to select the optimal coordinate axis for the database 11 to be searched while retaining the meaning of the coordinate axis.

  Next, the display unit 15 performs secondary display (step S25). That is, the music data is displayed based on the coordinate axis selected by the coordinate axis selection unit 14. This makes it possible to use a one-dimensional parameter by dimension compression of multi-dimensional feature quantities classified as easy to understand intuitively as a coordinate axis, and to select an optimal axis for music data in the database 11 for further display. Therefore, it is possible to realize the visualization of music data that does not depend on the target database and is easy to understand intuitively.

  FIG. 5 is a diagram showing a schematic configuration of a music search system to which the music visualization device of the present invention is applied. The input unit 31 receives an operation in which the system user selects data or a category displayed by the display unit 15 in the music visualization device 10 (or 20). For example, when the system is used with a pointer or a device having a touch panel, the user can directly touch the data display area for selection. An example of a user interface in the input unit 31 is shown in FIG. In FIG. 6, for example, the tempo of music and the timbre are selected as coordinate axes (features). The circles displayed on the coordinate axes individually indicate music data. Therefore, FIG. 6 shows a distribution of a plurality of music data on a two-dimensional coordinate determined by tempo and tone color. As shown in FIG. 6, the input unit 31 displays information on the categories classified by the classification processing unit 16 and a plurality of coordinate axis candidates, and the user can freely select from these. .

  The music reproducing unit 32 reproduces music data corresponding to the user input. Thereby, the user can audition any one of the music data obtained by the search.

  The characteristic operation of the present invention as described above is performed by causing a computer to execute a program. That is, the music visualization program of the present invention is a music visualization program for visually displaying the characteristics of music data, and reads out music data from a database and extracts a plurality of types of feature values from the read music data. A process of analyzing a distribution degree when each extracted feature quantity is represented on a one-dimensional coordinate axis, a process of selecting any one or more coordinate axes based on the distribution degree, and the selection A series of processes including a process of displaying the distribution of music data using the coordinate axes and the feature quantities corresponding to the coordinate axes are commanded to be readable and executable by a computer.

  In this manner, the degree of distribution when the feature values extracted from the music data are each represented on a one-dimensional coordinate axis is analyzed, and one or more coordinate axes are selected based on the degree of distribution, and the selected coordinate axis And the distribution of music data is displayed using the feature quantity corresponding to the coordinate axis, so the coordinate axis itself can be given meaning, and even if the music data stored in the database is biased, the meaning of the coordinate axis It is possible to select the optimum coordinate axis for the target database. In addition, the user can intuitively understand the distribution of music data. In addition, it is possible to shorten the time required for the user's music search and reduce the burden, and to promote understanding of the music stored in the database.

  The music visualization program of the present invention may further include a process of classifying music data stored in the database or each extracted feature and a process of displaying the classification result.

  As described above, since the music data stored in the database or each feature quantity extracted by the feature quantity extraction unit is classified, efficient music search can be performed even if the data quantity is enormous. .

  In the above-described embodiment, the music visualization device is shown as a functional block configuration, but even if expressed as a module, it is technically equivalent.

10, 20 Music Visualization Device 11 Database 12 Feature Amount Extraction Unit 13 Feature Amount Analysis Unit 14 Coordinate Axis Selection Unit 15 Display Unit 16 Classification Processing Unit 31 Input Unit 32 Music Playback Unit

Claims (7)

A music visualization device that visually displays the characteristics of music data,
A feature value extraction unit that reads out music data from a database and extracts a plurality of types of feature values from the read music data;
A feature quantity analysis unit for analyzing the degree of distribution when each extracted feature quantity is represented on a one-dimensional coordinate axis;
A coordinate axis selection unit that selects one or more coordinate axes based on the distribution degree;
A music visualization device comprising: a display unit that displays a distribution of music data using the selected coordinate axis and a feature amount corresponding to the coordinate axis. Further comprising a classification processing unit for classifying each feature quantity extracted by the music data stored in the database or the feature quantity extraction unit;
The music visualization device according to claim 1, wherein the display unit displays the classification result. The coordinate axis selection unit selects any of a plurality of coordinate axes as coordinate axis candidates based on the degree of distribution,
The music visualization according to claim 1, wherein the display unit displays a distribution of music data using a coordinate axis selected by a user operation and a feature amount corresponding to the coordinate axis. apparatus. A music visualization method for visually displaying the characteristics of music data,
In the feature quantity extraction unit, reading out music data from a database, and extracting a plurality of types of feature quantities from the read music data;
A step of analyzing the degree of distribution when the extracted feature quantities are each represented on a one-dimensional coordinate axis in the feature quantity analysis unit;
In the coordinate axis selection unit, selecting one or more coordinate axes based on the distribution degree;
And displaying the distribution of music data using the selected coordinate axis and the feature quantity corresponding to the coordinate axis in the display unit. In the classification processing unit, classifying the music data stored in the database or each feature amount extracted by the feature amount extraction unit;
The music visualization method according to claim 4, further comprising: displaying the classification result in the display unit. A music visualization program that visually displays the characteristics of music data,
A process of reading music data from the database and extracting a plurality of types of feature values from the read music data;
A process of analyzing the degree of distribution when each extracted feature quantity is represented on a one-dimensional coordinate axis;
A process of selecting any one or more coordinate axes based on the degree of distribution;
A series of processes including a process of displaying a distribution of music data using the selected coordinate axis and a feature amount corresponding to the coordinate axis are commanded to be readable and executable by a computer. Music visualization program. A process of classifying the music data stored in the database or the extracted feature quantities;
The music visualization program according to claim 6, further comprising: processing for displaying the classification result.

Images