JP2009042401A - Method for judging plagiarism of music piece - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a method for judging plagiarism of a music piece for automatically judging whether or not there is a plagiarizing or plagiarized relationship between two music pieces partially. <P>SOLUTION: The method includes: a digitized information extraction step for extracting note information of an arbitrary part of each music piece as digitized information that deviation from an average sound pitch is digitized in units of a prescribed sound length by one or a series of bars; a coincidence degree calculation step for comparing the digitized information of each music piece extracted by the digitized information extraction step and calculating the coincidence degree; and a judgment step for judging a plagiarizing or plagiarized relationship between two music pieces by comparing the coincidence degree with a predetermined reference value. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a technique useful for automatically determining whether a relationship of plagiarism / theft is established between two music pieces.

  In recent years, with the development of the Web, a site has been created where videos and music can be posted and viewed and viewed by an unspecified number of people. (You Tube, Yamaha Players Kingdom, etc.) However, postings by contributors who do not have copyright are a problem. In music, in addition to posting famous songs themselves, the act of plagiarizing and posting a part of them is also a problem. Therefore, at present, it is necessary for humans to determine whether or not the music posted on the side managing the posting site is plagiarized. This is unrealistic because it takes a lot of time and manpower.

  On the other hand, various studies have been conducted on the similarity determination between two music pieces. This is a technique for automatically extracting a similar song from many songs when a favorite song is designated when searching for a favorite song of a searcher. Therefore, the similarity determination is aimed at determining whether or not the impression is similar for the entire music.

Also, in determining the similarity between any two melodies on the musical score information, the original music written in an arbitrary key was transposed to the same key, and the memory effect, which is a human audible characteristic, was applied. A method of obtaining a determination result based on the real sense by comparing the three elements of the tone string / pitch time (auditory sound length) / sounding point and determining the similarity of the melody is also considered.
JP 2001-265324 A

  However, although plagiarism is plagiarism even if it partially matches, according to the former configuration, it is difficult to determine whether or not it is plagiarism because only the impression of the entire song is determined.

  Further, according to the latter configuration, the determination is made based on whether or not the pitches are completely coincident. However, the difference between the semitone difference and the three tone difference is different. Moreover, since all of them are transposed to a certain key and compared, local #, ♭, and natural symbols cannot be used, and most of them have a single tone difference and similar similarity, such as a local semitone difference. Is difficult to detect.

  In other words, it is possible to apply a key signature that converts only the plagiarism part, but if the key is converted locally, the conventional judgment method that transposes the whole could not detect the plagiarism. . In addition, it is highly likely that a piece of music having the same key and different average pitches and the same deviation from the same, for example, the same C key and the dredredre and the remyremy are not plagiarism in the prior art.

  The present invention has been made in view of such circumstances, and provides a music plagiarism determination method that automatically determines whether or not a relation of plagiarism / non- plagiarism is partially established for two music pieces. The main issue.

  In order to achieve the above object, a music plagiarism determination method according to the present invention is a music plagiarism determination method for determining whether or not a relation of plagiarism / theft is established between two music pieces. A digitized information extraction step for extracting note information of an arbitrary part of each musical piece in a unit of measure as digitized information obtained by digitizing a deviation from an average pitch in a predetermined pitch unit, and the digitized information extracting step A degree of coincidence calculating step for comparing the digitized information of an arbitrary part of each extracted piece of music and calculating the degree of coincidence; and comparing the degree of coincidence with a predetermined reference value for plagiarism between two pieces of music; And a determination step of determining the relationship of theft.

  Therefore, in the digitized information extraction step, the note information of an arbitrary part of each piece of music on the plagiarism side and the plagiarism side in units of one or a plurality of consecutive measures is numerically expressed as a deviation from the average pitch in a predetermined pitch unit. In the coincidence degree calculation step, the degree of coincidence is calculated by comparing the digitized information of an arbitrary part of each song. In the determination step, the degree of coincidence is compared with a predetermined reference value. Since the relationship between plagiarism / theft from two music pieces is determined, it is possible to automatically determine whether or not there is a partial match between the music pieces. In addition, since the degree of coincidence is calculated by comparing the digitized information obtained by quantifying the deviation from the average pitch, there is no need to transpose the music to be compared, and erroneous determination associated with transposition is eliminated. Is possible.

  Here, the degree of coincidence calculation step may calculate the degree of coincidence by integrating the numerical value differences for each predetermined sound length unit based on the digitized information of each piece of music to be compared. For example, the calculation may be performed by integrating the squares of the pitch differences for each predetermined pitch unit.

  Further, the digitized information extraction step specifically includes a pitch conversion step for dividing note information of an arbitrary part of the music piece into predetermined pitch units and converting them into numerical values representing respective pitches, and the pitch Based on the numerical value converted in the conversion step, an average pitch calculation step for calculating an average pitch in the one or a plurality of continuous measures, and the pitch conversion in the one or a plurality of continuous measures A relative pitch calculation step of calculating a difference from the average pitch converted by the step for each predetermined pitch unit may be configured.

  In addition, in order to improve the determination accuracy, it is preferable to use a sixteenth note length or a thirty second note length as the above-mentioned predetermined sound length unit.

  As described above, according to the present invention, the musical note information of an arbitrary part of each piece of music in units of one or a plurality of continuous measures is digitized by quantifying the deviation from the average pitch in a predetermined pitch unit. Information is extracted, and the degree of matching is calculated by comparing the digitized information of an arbitrary part of each extracted piece of music, and the degree of matching is compared with a predetermined reference value to obtain plagiarism between two pieces of music.・ Since the relationship between plagiarism is determined, a method for determining the plagiarism that can automatically determine whether or not the relationship between plagiarism and non-piracy is partially established for two songs is provided. It becomes possible.

  In addition, note information is extracted as digitized information obtained by quantifying the deviation from the average pitch in a predetermined pitch unit, so the degree of coincidence is calculated by the relative pitch difference for each comparison unit. And transposition is unnecessary. For this reason, it is possible to effectively detect a local transposition or a case where there is a similarity such as a difference in semitones in almost a single tone.

  The best mode for carrying out the invention will be described below with reference to the drawings.

  The two songs are divided into a plagiarism side and a plagiarism side. In the present invention, a musical piece on the theft side is prepared in advance. Then, it is determined whether or not another song is a plagiarism of a song prepared in advance. Also, the plagiarism of music is plagiarism if a part of the music matches. FIG. 1 shows music scores of music A and music B. Here, the parts of the second bar and the third bar of the music A coincide with the third bar and the fourth bar of the music B. In this way, plagiarism is established even if the place in the whole song is different.

  Also, as shown in FIG. 2, when the music is transposed, the relative change is the same even if the absolute value of the pitch is different, and this also establishes the plagiarism relationship. Even if there is no modulation, by using local #, ♭, natural symbols, etc., it is possible to easily create a music piece that is different in semitones, in one tone, or in two tones. Further, without using the modulation or the above-mentioned local #, ♭, and natural symbols, for example, it is possible to easily create, for example, a music that is mostly different by one tone and only partially different by semitones. Also, as shown in FIG. 3, plagiarism is suspected even if there is a part where the rest position is different in addition to the modulation.

  The present invention provides a method for accurately determining plagiarism and non-piracy between the above-described music pieces. In the present embodiment, an example shown in FIG. 3 will be mainly described.

First, a configuration example of a music plagiarism determination system to be used is shown in FIG.
This music plagiarism determination system is constituted by a computer as an information processing apparatus having a CPU, ROM, RAM, etc., and performs predetermined processing by a predetermined program held in a memory. It is connected to an input device (not shown) such as a CD that inputs music information and music information subject to plagiarism determination), and an output device (not shown) such as a monitor or printer that outputs the result of determination of plagiarism.

  This computer automatically digitizes the melody information of music input as analog signals (melody information of music subject to plagiarism judgment and melodic information of music subject to plagiarism judgment) to digitize the note information Unit 1, segment division unit 2 that performs segment division of the music that is the target of the plagiarism determination and segment division of the musical piece that is the target of the plagiarism determination, and a rest that removes rest information from the input music information The note information removing unit 3, the 16th note converting unit 4 for dividing the note information from which the rest information has been removed into 16th note units and converting them into numerical values representing the respective pitches, and the average sound in the divided units An average pitch calculation unit 5 that calculates a pitch, and a relative pitch calculation unit 6 that calculates a difference (relative pitch) between a numerical value representing a pitch converted into a sixteenth tone and an average pitch in divided units. And divided segmental information on plagiarism determination target music Information holding section 7 for each section for the plagiarism determination target music to be held, section information holding section 8 for each section for the plagiarism determination target music for storing the divided section information of the music for determination of theft plagiarism, and the section for the plagiarism determination target music And a section combination control unit 9 that controls the combination of the section of the pirated judgment target music, the section information for the pirated judgment target music, and the section specific information for the pirated judgment target music combined by the combination control unit, The relative relative pitch residual average calculating unit 10 that calculates the average of the relative pitch residuals by zone based on the above, and the relative pitch residuals by zone calculated by the relative relative pitch residual average calculating unit 10 by the zone And a plagiarism-related determination unit 11 that determines the presence or absence of plagiarism based on the average.

  In such a configuration, first, when the melody information of the music to be determined for plagiarism and the melody information of the music to be determined for plagiarism are input to the system, the melody information of the music input by the digitization processing unit 1 is obtained. Automatically record notes and digitize note information. However, if the input is already digitized information, this processing unit is bypassed (or not installed from the beginning). The digitized melody information is divided into predetermined sections (for example, in bar units) by the section dividing section 2, and after the rest information is removed by the rest information removing section 3, a sixteenth note conversion section is obtained. 3, the note information is divided into sixteenth note lengths and converted into numerical values representing the respective pitches. The segment division process may be performed after the rest removal process. However, if the segment is divided after the rest is removed, the method for identifying the division part becomes cumbersome. It is preferably performed before the removal process.

  Thereafter, an average pitch is calculated for each section divided by the average pitch calculation unit 5, and a pitch variation based on the average pitch for each section divided by the relative pitch calculation unit 6, that is, 16 A difference (relative pitch) between a numerical value representing a pitch converted into a partial sound and an average pitch is calculated. Since the music to be processed here is the plagiarism determination target music and the stolen determination target music, the information regarding the relative pitch of each divided section of the plagiarism determination target music calculated by the relative pitch calculation unit 6 is the plagiarism determination target. Information related to the relative pitch of each divided section of the musical composition subject to theft determination determined by the relative pitch calculation section 6 and held in the musical section specific information storage unit 7 is information related to the stolen determination target musical section. It is held by the holding unit 8.

  Information about how many bars each piece of music information is divided is sent from the section dividing unit 4 to the target section combination control unit 9. Thereafter, the section-specific information holding section 7 for the plagiarism determination target music, the theft detection determination so as to calculate a residual average of relative pitches for all combinations of the divided sections of the respective music from the target section combination control section 9 Instructions are sent to the section-by-section information holding section 8 for the target music and the relative pitch residual average calculating section 10 by section. The average relative pitch residual difference calculated by this is sent to the plagiarism relationship determination unit 11, where the presence / absence of plagiarism is determined and the result is output.

FIG. 5 shows a flowchart of the determination method using the system configuration described above, and the following description will be given based on this flowchart.
First, the musical score information of the two pieces of music shown in FIG. 3, that is, the melodic information for the plagiarism determination target music (musical score information) and the melodic information for the plagiarism determination target music (musical score information) are input in a digitized state. (Step S1). Thereafter, a combination of consecutive n bars of each music is extracted (n is 1 or more). (m−n + 1) combinations are extracted from the music with m measures (step S2). This series of n novels is a comparison unit for comparing music pieces.

Then, rest information is removed for each of the divided sections (step S3), and the remaining note information is expressed numerically in terms of a 16th note, which is the minimum length on the score (step S4). For example, in the musical scores of A and B in FIG. 3, if the section to be divided (comparison unit) is converted as a bar unit, each can be expressed as follows. Here, () represents one measure. Each numerical value is a relative pitch (half-tone difference is set to 1) when the pitch of the C-tone is 0. At this stage, the rest information has already been removed.
A1 ... (7,7,7,7,7,7,7,7,9,9,9,9,7,7,7,7)
(9, 9, 9, 9, 9, 9, 9, 9, 11, 11, 12, 12, 9, 9, 9, 9)
(11, 11, 11, 11, 11, 11, 11, 11, 14, 14, 14, 14)
(14, 14, 14, 14, 11, 11, 9, 9, 7, 7, 7, 7)
B1 ... (7, 7, 7, 7, 7, 7, 7, 7, 10, 10, 10, 10)
(10, 10, 10, 10, 10, 10, 10, 10, 8, 8, 8, 8, 10, 10, 10, 10)
(10, 10, 10, 10, 7, 7, 5, 5, 3, 3, 3, 3)
(3, 3, 3, 3, 2, 2, 2, 2, 3, 3, 3, 3)

Next, based on this, the average pitch of each measure of A1 and B1 is calculated (step S5). The average pitch for each measure of A1 and B1 is as follows.
Average pitch of each measure of A1
First measure 7.5
Second measure 9.625
Measure 3 12.0
Measure 4.333
Average pitch of each measure of Bl
First measure 8.0
Second measure 9.5
3rd measure 6.3333
Measure 4.667

Then, a difference (relative pitch) between each note and the average pitch is calculated for each value and each measure (step S6). The results are as A2 and B2, respectively.
A2 (-0.5, -0.5, -0.5, -0.5, -0.5, -0.5, -0.5, -0.5, 1.5, 1.5) ,
1.5, 1.5, -0.5, -0.5, -0.5)
(-0.625, -0.625, -0.625, -0.625, -0.625, -0.625, -0.625, -0.625
1.375, 1.375, 2.375, 2.375, -0.625, -0.625, -0.625, -0.625),
(-1, -1, -1, -1, -1, -1, -1, -1, 2, 2, 2, 2),
(3.667, 3.667, 3.667, 3.667, 0.667, 0.667, -1.333, -1.333,
-3.333, -3.333, -3.333, -3.333)
B2 ... (-1, -1, -1, -1, -1, -1, -1, -1, 2, 2, 2, 2),
(0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, 0.5, -1.5, -1.5, -l.5,
-1.5, 0.5, 0.5, 0.5, 0.5)
(3.667, 3.667, 3.667, 3.667, 0.667, 0.667, -1.333, -1.333,
-3.333, -3.333, -3.333, -3.333),
(0.333, 0.333, 0.33, 0.333, -0.667, -0.667, -0.667, -0.667, 0.33, 0.33, 0.333, 0 .333)

  Thereafter, all combinations of consecutive pairs of n bars extracted by the two music pieces are extracted (step S7). In the example of FIG. 3, if the division unit (comparison unit) is set to one measure, there are 4 measures each of the music for determination of plagiarism and the music for determination of theft, and the combination of the divided segments in both music is 4 ×. 4 to 16 combinations.

Thereafter, the average value of the squares of the relative pitch differences is calculated for all the extracted combinations (step S8). That is, the average value _{Zij of the} squares of the relative pitch differences given by the following formula (1) is calculated based on the relative pitches of A2 and B2 in the combined measures.

  Here, i and j indicate the number of measures of A2 and B2, respectively, k indicates the number of measures when one measure is divided into 16th notes, and A2ik and B2jk are the i numbers of A2 and B2, respectively. , The kth pitch in 16th note conversion of the jth measure. N is the number of notes in 16th note conversion to be compared. Assuming that the unit to be compared is one measure, a measure having no rest is N = 16, and a measure having a rest is a value obtained by subtracting the rest from 16.

When the above-described value Z _ij is smaller than a predetermined value, it is determined that there is a plagiarism relationship between the two songs, and otherwise, it is determined that there is no plagiarism relationship (step S9).
For example, if Z _ij is calculated with the above A2 and B2, Z ₃₁ = 0 and Z ₄₃ = 0, and the third measure of A2 and the first measure of B2 and the fourth measure of A2 and the third measure of B2 are It is determined that there is a plagiarism relationship that matches completely.

  Therefore, according to the configuration as described above, note information of an arbitrary part of each piece of music on the plagiarism side and the plagiarism side is deviated from the average pitch in a predetermined pitch unit in units of one or a plurality of continuous measures. Is extracted as digitized information, and the degree of coincidence is calculated by comparing the digitized information of an arbitrary part of each extracted piece of music, and this degree of coincidence is compared with a predetermined reference value, Since the relationship between plagiarism / theft from music is determined, it is possible to automatically determine whether the relationship between the plagiarism / non-theft is partially established for two music pieces.

  In addition, note information is extracted as digitized information obtained by quantifying the deviation from the average pitch in a predetermined pitch unit, so the degree of coincidence is calculated by the relative pitch difference for each comparison unit. Therefore, the transposition operation becomes unnecessary. For this reason, it is possible to determine effectively when there is a local transposition or when there is a similarity such as a difference of almost one tone and a difference of semitones locally.

  In the configuration described above, the comparison unit is set for each bar, but is not limited to this, and a plurality of continuous bars, such as every two consecutive bars, every four consecutive bars, etc. A comparison unit may be used, in which case N is twice or four times as large as when N is one bar. In that case, if divided into consecutive n bars, the music of m bars as a whole is divided into mn + 1 sections. (When dividing into two consecutive bars, it is divided into sections such as first and second bars, second and third bars, and third and fourth bars.)

  In the bar from which the rest is removed, there is a difference in the number of relative pitches to be compared. In this case, the portion to be compared with the rest portion is assumed to have a certain pitch difference. You may make it process.

  Furthermore, in the above description, it is determined that there is a plagiarism relationship when Z = 0, but even if Z = 0, it may be determined that there is a plagiarism relationship if it is equal to or less than a predetermined value. it can. It is also possible to set the reference value in multiple stages and to give a wide range of judgment results (for example, “suspected plagiarism is strong”, “suspected plagiarism cannot be denied”, “no plagiarism is suspected”) Etc.)

Further, in the above-described configuration, the method of integrating the numerical difference for each predetermined pitch unit is calculated by the mean value Z _{ij of the} square of the pitch difference given by the equation (1). For example, you may make it substitute with the residual average type | formula etc. which are shown to following Numerical formula (2), Numerical formula (3), Numerical formula (4).

  In the above-described configuration, the example in which the melody information of the music to be pirated is also input together with the melody information of the music to be pirated is shown. The database 11 may be provided, and the melody data of the stolen work determination target music to be compared may be extracted from the database, and the same processing may be performed.

  Furthermore, in the above-described configuration, an example in which the note information of each piece of music is decomposed into sixteenth note lengths is shown, but the present invention is not limited to this. You may make it perform.

FIG. 1 is a diagram illustrating an example of musical score of a musical piece A on the plagiarism side and a musical piece B on the plagiarism side. FIG. 2 is a diagram illustrating an example of musical score of the musical piece A on the plagiarism side and the musical piece B on the plagiarism side. FIG. 3 is a diagram showing an example of musical score of the musical piece A on the plagiarism side and the musical piece B on the plagiarism side. FIG. 4 is a functional block diagram illustrating a configuration example of the music plagiarism determination system. FIG. 5 is a flowchart showing a music plagiarism determination method.

Claims (4)

In the music plagiarism determination method for determining whether the relationship of plagiarism / theft is established between two music pieces,
A digitized information extraction step of extracting note information of an arbitrary part of each musical piece in units of one or a plurality of continuous bars as digitized information obtained by digitizing a deviation from an average pitch in a predetermined pitch unit;
A degree of coincidence calculation step for comparing the digitized information of an arbitrary part of each piece of music extracted by the digitized information extraction step and calculating the degree of coincidence;
And a determination step of determining the relationship between plagiarism and plagiarism between two pieces of music by comparing the degree of coincidence with a predetermined reference value. 2. The coincidence degree calculating step calculates the degree of coincidence by integrating the numerical difference for each predetermined sound length unit based on the digitized information of each piece of music to be compared. No music plagiarism determination method. The digitized information extraction step
A pitch conversion step for dividing note information of an arbitrary part of the music piece into predetermined pitch units and converting them into numerical values representing respective pitches;
Based on the numerical value converted in the pitch conversion step, an average pitch calculation step for calculating an average pitch in the one or a plurality of continuous measures;
A relative pitch calculating step of calculating, for each predetermined pitch unit, a difference from the average pitch of the numerical value converted by the pitch converting step in the one or a plurality of consecutive bar units. 2. The method for determining the plagiarism of music according to claim 1. The music plagiarism determination method according to claim 1, wherein the predetermined tone length unit is a sixteenth note length or a thirty-second note length.

Images