JP2797317B2 - Automatic composer - Google Patents

Description

Description: TECHNICAL FIELD OF THE INVENTION The present invention relates to an automatic composer.

[Background] One of the issues required for an automatic composer is the ability to efficiently generate a song as required by a user.

Here, considering the number of melody combinations, if the number of pitches that can be taken by each sound of the melody is x and the number of notes is y, the number of combinations is very much ^xy, even in the pitch sequence alone. A huge number. Therefore, an automatic composer that generates the pitch of each sound of the melody with uniform probability (generates a random number) can be easily realized, but it becomes more distant until a melody desired by the user is obtained. Must be repeated, and cannot be used practically.

Therefore, the applicant of the present application has proposed a melody feature parameter (melody control parameter) for characterizing a melody to be generated.
Introducing the concept of
There has been proposed an automatic composer that limits a certain musical knowledge (for example, adjacent pitch difference is within one octave at most) and generates a melody based on the limited melody control parameter and chord progression. No.62-145405
issue). In the automatic music composer of this application, the melody control parameter can take any value within the range of the limitation by music knowledge, and when the judgment result of the user for the generated melody is negative, the melody control parameter is changed. The melody is generated again. However, there is no guarantee that the regenerated melody will be a melody that is closer to the user's request than before, and the problem of the ability to generate a melody that meets the user's preference remains.

[Object of the Invention] The present invention is a further improvement of the automatic composer of the above-mentioned application, and an object of the present invention is to provide an automatic composer capable of more effectively generating a melody that suits individual users' preferences. It is to be.

[Gist of the Invention] In order to achieve the above object, the present invention focuses on notifying a user's preference to an automatic composer as basic information for composing, and the result of judgment on the generated melody (user's (Preference) is stored as frequency data on which information called a melody control parameter is generated.

Claim 1 is a melody control parameter generation means for generating a melody control parameter, and a melody synthesis for synthesizing a melody based on the melody control parameter generated thereby and a given chord progression. The present invention is applied to an automatic music composer of the type having means. In the configuration of claim 1, the frequency data stored in the frequency data storage means serves as an index when the melody control parameter generation means generates a melody control parameter. Specifically, the frequency data indicates the frequency at which each value of each melody control parameter is generated. Here, this frequency reflects the user's judgment result evaluated for the melody synthesized by the melody synthesis generating means. The reason is that each time the user's judgment result for the synthesized melody output by the output means is input, the frequency data changing means is activated, and the frequency data in the frequency data storage means is output according to the judgment result. This is because the part related to the value of the melody control parameter used for the synthesis of the melody is changed. More specifically, the frequency data changing means determines whether a user likes or dislikes the result of the user's judgment.
Even when the type information is given, after obtaining the value of the melody control parameter for the output melody, the value of the frequency data relating to the melody control parameter for that value is increased for “likes” and for “dislikes” Decrease. Therefore, when a new melody is subsequently generated, each control parameter generated by the melody control parameter generating means can easily take a value that more closely matches the user's preference according to the frequency data. For this reason, the melody generated by the melody synthesizing means reflects the user's preference, and the intended purpose can be achieved.

Claim 2 applies the present invention to an automatic composer that composes music based on a melody input by a user. According to claim 2, melody control parameter generation means for generating a plurality of types of melody control parameters, and melody synthesis means for synthesizing a melody based on the melody control parameters generated by the melody control parameters and a given chord progression And a frequency data storage means for storing the frequency data for controlling the frequency at which various melody control parameters occur at specific values by type and by value, and the melody synthesizing means. Output means for outputting a melody; input means for inputting a user's judgment result on the melody output by the output means; and melody output by the output means among the frequency data stored in the storage means. For each melody control parameter value used for synthesis, Frequency data changing means for changing the assigned frequency data in accordance with the judgment result input by the input means, wherein the melody control parameter generating means newly generates the melody control parameter based on the frequency data changed by the frequency data changing means. An automatic composer characterized by generating a melody control parameter for controlling a melody to be generated, wherein the melody synthesizing means synthesizes a melody with the generated melody control parameter,
In addition to the above configuration, melody input means for inputting a melody that forms a part of a song, and input melody analysis means for analyzing the input melody are provided, and the analysis result of the input melody analysis means is further included in the melody control parameter generation means. Means for generating a melody control parameter based on the According to this configuration, the entire music is divided into a plurality of sections, a melody is generated by obtaining a melody control parameter from frequency data for a certain section, and a melody is obtained by obtaining a melody control parameter from an input melody for a certain section. Can be generated, and the characteristics of the input melody, which is a direct reflection of the user's preferences, can be incorporated into the generated melody, and the melody can be generated based on the frequency data incorporating the user's preferences. .

Claim 3 relates to a configuration example for obtaining a value of a melody control parameter defining a change target when changing frequency data. According to this, the frequency data changing means has melody analysis means for analyzing the melody output by the output means.

This configuration is particularly effective in the case of an automatic composer of a type having input melody analysis means for analyzing the input melody in order to compose music based on the input melody, and this input melody analysis means The output melody can also be used as analysis means.

Of course, the value of the melody control parameter that determines which frequency data is to be changed, that is, the value of the melody control parameter used for synthesizing the output melody can be obtained without the melody analysis means. For example, the value of the generated melody feature parameter may be temporarily stored during operation of the melody feature parameter generation unit, and the value may be referred to at the time of operation of the frequency data change unit to determine the frequency data change target. .

An embodiment of the present invention will be described below with reference to the drawings.

Overall Configuration FIG. 1 shows the overall functions of the automatic composer according to the present embodiment. In the figure, a melody generation control parameter generation unit 1 generates a melody control parameter (melody feature parameter) according to frequency data stored in a frequency data storage unit 9. The melody control parameters are sent to the melody synthesizing means 3 to control the generation of the melody. The melody synthesizing unit 3 synthesizes a melody based on the melody control parameter from the melody generation control parameter generating unit 1 and the chord progression stored in the chord progression storage unit 2 and stores the melody in the melody storage unit 4. The melody stored in the melody storage means 4 is monitored by the monitor means 5 as needed. The user makes a judgment as to whether he / she likes or dislikes the monitored melody, and inputs the judgment result to the input means 6. On the other hand, the melody analysis means 7 operates to analyze the melody determined by the user and extract the value of the melody control parameter used for synthesizing the melody. Then, the data changing means 8 operates, and changes the frequency data for the other melody control parameters, which are the analysis results of the melody analyzing means 7, according to the judgment result input from the input means 6. Thus, the content of the frequency data storage means 9 is updated. When a melody is later synthesized, the melody generation control parameter generation means 1 generates a melody control parameter in accordance with the updated frequency data, so that the generated melody reflects the preference of the user more.

FIG. 2 shows a hardware configuration example for realizing the automatic music composition machine shown in the functional block diagram in FIG.

In FIG. 2, the CPU 10 implements melody analysis, synthesis, and frequency data change functions. Work memory
Reference numeral 20 denotes a memory for temporarily storing variables and the like used by the CPU 10 during operation. Frequency data storage device 30, input device 4
The monitor 50 corresponds to the frequency data storage means 9, the monitor means 5, and the input means 6 in FIG. The monitor 50 includes a display device such as a CRT, a sound source and a sound output device, a score printer, and the like.

FIG. 3 shows a main flow of the operation of the present embodiment. 3-
At step 1, the area allocation of the work memory 20, etc. is performed, and
At 2 the frequency data stored in the frequency data storage device 30 is read into the work memory 20, and at 3-3 a key input from the user is waited. Each time there is a key input (3-4 to
3-7) and execute the corresponding processing (3-8 to 3-11). For example, when a composition instruction is given, a melody is generated based on frequency data. (3-4, 3-8). Thereafter, when there is a monitor request, the monitor is executed (3-9 to 3-11). When the judgment on the monitored melody is input (3-5, 3-6), the frequency data is changed according to the judgment result (3-9, 3-1).
0).

Variable List and Data Format FIGS. 4 to 7 show a list of main variables used in the flowcharts described later and their data formats. Each of them will be referred to when describing a flowchart described later, and the description will be omitted here. However,
The frequency data related to the features of the present invention will be described here.

The frequency data DATAij is stored in the frequency data storage device 30 as shown in FIG. Where DATAij is the first subscript i
Represents the type of control parameter (for example, the number of harmony sounds, the number of non-harmonic sounds, etc.), the subscript j represents the value of the parameter of type i, and the value of DATAij itself is the value of the parameter of type i. i represents the frequency. Therefore, when a feature parameter is generated during melody synthesis, the value of the control parameter i is most likely to be the value j indicated by the highest value DATAij among DATAij (j = 1 to N). Here, the entire frequency data {DATAij} reflects the user's judgment on the melody, so that a melody control parameter more suited to the user's preference, that is, a melody is more easily generated. In the format (contents) shown in FIG. 7, the total number of parameter types stored in the frequency data storage device 30 is placed at address 0, and DATA ₁ , ₁ , DATA _{1, e} , and DATA ₂ are stored at subsequent addresses after ₁ . _{, 1} , DA
The addresses TA _{2, 2,} ... are stored. And DATA
_From the address of _{i, 1} (address of data corresponding to the value 1 of the parameter of type i) to the address of DATA _{i, e} (type i
, DATA _i , DATA _i, ... DATA _{1, e} are successively stored.

Hereinafter, individual processes in the main flow (FIG. 3) will be described in more detail.

Reading of frequency data 7 FIG. 8 shows a flow of reading frequency data executed in 2-2 of the main flow. At 8-1, the number of parameter types is read and set to N, and at 8-2, the type is set to i. In a loop from 8-3 to 8-10, for each type i from 1 to N, (8-9, 8-10),
The start address s and the end address e of the frequency data of that type are read (8-3, 8-4), and j is changed from 1 to e−S +
By moving to 1 (8-5, 8-7, 8-8)
The address is moved from the start address s to the end address e, and the data DATAij of each address is read.
(8-6).

Melody generation FIG. 9 shows a flow for melody generation 3-8 in the main flow (FIG. 3).

A control parameter is generated in 9-1, a melody of a harmony is generated in 9-2, a non-harmonic is added to a melody of a harmony in 9-3, and a data format of the melody is exchanged in 9-4. I have. These processes are performed in units of one section (one bar).

Here, the generation of the control parameter 9-1 is performed based on the frequency data stored in the data storage device 30. Specifically, for example, the value of (DATAij + random element) is calculated from j = 1 to a predetermined value (for all possible values of the parameter i), and the value of j at the maximum value is identified as i. Is adopted as the value of the parameter. This process is repeated for all parameter types (i = 1 to N). If there is no random element, a value j corresponding to the highest frequency data is generated as the value of each type of parameter among the frequency data of each type. Other processing in melody generation 9 2-2 to 9-4 will be described below. In the drawings referred to in the description, many melody control parameters (for example, RRAR
P, RRNST, LL ...), but it is not necessary to prepare frequency data for all these types,
Certain control parameters can be derived from basic control parameters. For example, if the number of harmony sounds is directly obtained from the frequency data, the generation timing information RRART of the harmony sound is obtained from the number of harmony sounds according to Patent Application No. 62-1.
It can be generated by a note dividing / combining means using a pulse scale as shown in 21037.

<Generation of harmony> FIG. 10 shows a flow of harmony generation performed in melody generation (FIG. 9).

In Fig. 10, to generate the harmony pattern {LL},
Features of the harmony pattern PC ₂ to PC ₅ are used. PC
₂ is the maximum value (highest tone) among LL ₁ to LL _N (N is the number of generated chords), PC ₃ is the minimum value of LL ₁ to LL _N , and PC ₄ is | LL ₁ -LL ₂ | ~ | L
The maximum value PC ₅ in L _N−1 LL _N | limits the minimum value in | LL ₁ −LL ₂ |.

The purpose of the harmony generation in FIG. 10 is to generate a harmony pattern to be included in the generated melody.

More specifically, at 10-1, the counter i of the basic unit length in the bar is initialized to 1, and the generated harmony (precisely, LL
) Is initialized to 1. It is determined whether or not the timing i of interest in 10-2 is the timing of the generation of the harmony sound, by the bit i of the harmony generation timing information RRARR
Is determined by examining. The current position (measure number × BEAT) of interest at 10-3 + i (where BEAT is 1
The number of the chord corresponding to the number of the basic unit time constituting the bar) is obtained from the sequence of the chord length of the chord progression to obtain the number CKNO of constituent notes of the chord CDi. The number of constituent sounds is obtained because the octave number and the number of chord constituent sounds within one octave are defined as the data format of the harmony pattern {LL} (see FIG. 6). Next, a parameter PC ₂ that limit the upper and lower limits of harmonic tones pattern respectively 10-4, and a PC ₃ and coding scheme number of sounds CKNO, random manner to produce a candidate LL _NO element of harmonic tones pattern. The first pattern element is determined by this (10-5, 10-8)
(See Reference.) For the second and subsequent pattern elements, the difference from the immediately preceding pattern element needs to be within a limited range. Therefore, in 10-6, parameters PC ₄ and P ₄ that limit the maximum value and the minimum value of the difference between adjacent pattern elements, respectively.
To check the authenticity of the LL _NO with C _5. If the result of the check is not pass (OK) (10-7), the processing of 10-4 and below is repeated again. Depending on the value of PC, the following LL _NO
If the process is repeated a predetermined number of times, it may happen that the conditions of the parameters PC ₄ and PC ₅ are never satisfied. PC
_{4. It} is necessary to give up the restrictions imposed by PC ₅ and break out of the loop. (Not shown).

Once one pattern element is determined, the counter for the harmony
No is incremented (10-8), the timing i is advanced to the next position (10-9), and the processing of 10-2 or less is repeated from i = 1 to BEAT.

This is a pattern {LL} generation of harmonic sounds as, at each occurrence position indicated by the generation timing information harmonic sounds, successively forming a pattern element from the control parameter PC ₂ to PC _5.

FIG. 11 shows a flow for converting the obtained harmony pattern {LL} pitch, that is, the melody of the harmony {MD}, obtained as described above.

First, in step 11-1, the counter llc, which is the number of the harmony generated in the bar, is initialized to 0, and the timing i of the basic unit length in the bar is set to 1. 11-2
Referring to bit i of the timing information of the generation of the harmony,
It is determined whether or not the current position is the timing at which a harmony occurs.
If not, move the current position to the next (11
-5) The inspection is performed again, and if the position is a generation position, the process goes to 11-3 to increment the harmony counter llc. Then, the code CDc corresponding to the position of interest (measure number × BEAT) + i is determined from the code length sequence {CRi} of the chord progression (11-4). Next, the constituent sound data C of the code CDc
C (each digit of 12 bits indicates the position of the pitch name, and bit “1” indicates that the pitch name is used as a chord constituent) (11-5), and sets the counter nc of the chord configuration sound to 0 (11-6), the note name counter j is set to 1 indicating C (11-7), and then moved in the range of j = 1 to 12 (11-2,
11-3) In the meantime, every time bit j = 1 of CC is found, nc is incremented (11-8, 11-9), and the constituent tone number nc at that time is the corresponding element LLLlc of the non-harmonic tone pattern.
It is checked whether it matches the constituent note name in the lower digit (11-10). If they match in 11-10, pitch data MDlli is generated with j at that time as the pitch name and the data in the upper digit of LLllc as the octave number (11-11). By repeating this process from i = 1 to BEAT, (11-1
4) The harmony pattern {LL} for one bar is exchanged for the melody {MD} of the harmony.

<Addition of non-harmonic sound> FIG. 12 shows a flowchart of non-harmonic sound generation executed in step 9-3 of melody generation (FIG. 9). The purpose of this processing is to use the melody of the harmony {MD}, the harmony generation timing information RRARP and the non-harmony generation timing information RRNCT to add the non-harmony melody ND To generate non-harmonic pitch data. MDi and NDi have pitch information, MDi and R
The relationship with RART is determined by the pitch MDi of the i-th chord in the measure at the timing of "1" that appears i-th from the last digit of the RRART.
And the relationship between NDi and RRNCT is the same. (See FIG. 13).

According to the flow of FIG. 12, the counter i of the basic unit tone length is set to "1" in 12-1, and the counter arpc of the bit number which is "1" of the harmony generation timing information RRARP is set to "0". Set to “1” in the non-harmonic generation timing information RRNCT
The counter rsc of the number of bits is set to 0.

At 12-2, it is determined whether the bit i of the RRARP is "1" or i is "1", and if it is established, as pre-processing, the situation before and after the position of interest is examined. Function indicating
f ₁ and f ₂ are determined, the number of non-harmonic sounds inrsc sandwiched between the harmonious sounds (or bar lines) is calculated, and the position of each non-harmonic sound is determined as a stack J. To be stored. 12−
When the number of non-harmonic tones obtained in step 3 is 1 (12-4), a process of adding one non-harmonic tone is executed (12-5). (12-6, 12-7). If it is not successful, the position stored in J is used as the non-chord generation timing information RRNC.
Execute the process to delete from T. (12-6, 12-8).
When the number inrsc to be added is 2, a process of adding two non-harmonic tones is executed (12-4, 12-9), and when the addition is successful, each pitch md1, md2 is pushed to the ND. (12-10, 1
2-11). In the case of failure, the position to be added stored in J is used as the non-harmonic generation timing information RRNCT.
Remove from. (12-10, 12-12). In either case,
Move the timing i to the next position (12-3), i = BEAT
Until the above, the processing of 12-2 and below is repeated. In FIG. 12, it is assumed that at most two non-harmonic sounds are added between the harmonic sounds, but the flow may be modified so that three or more non-harmonic sounds can be added. In order to ensure that at most two non-harmonic sounds are attached between harmony sounds, the harmony generation timing information RRART and the non-harmony generation timing information RRNCT must be set in advance.
In comparison with the above, only two non-harmonic sound generation timings may be provided between adjacent harmonic sound generation timings (when generating non-harmonic sound generation timing information).

Details of 12-3, 12-5, and 12-9 in FIG.
These are shown in FIG. 14, FIG. 15, and FIG.

In the preprocessing (FIG. 14), one or two non-harmonic sounds of interest are examined before and after the position to determine the functions f ₁ and f _2, and the previous harmonic sound (or bar line) is determined. Number of non-harmonics sandwiched between the following harmonic (or bar line) in
The rsc and the position J of each non-harmonic tone are obtained. The function f ₁ becomes “1” when there is a harmony in front of the position of the non-harmonic to be focused in the bar, and becomes “−1” when there is no harmony, that is, when the front is a bar line. . The function f ₂ is “1” when there is a harmony behind the position of the non-harmony in the bar.
Otherwise, it is "-1".

According to the flow shown in FIG. 14, it is first checked at 14-1 whether the current position is at the beginning of a bar (i = 1) and not at the timing of generation of a harmony sound (bit i of RRARP is "0"). Sometimes, the start position inext at the time of searching for a harmony behind the current position i is set to i, and a function f ₁ indicating the situation in front is set to “−1” indicating a bar line, and The counter inrsc for counting the number of non-harmonic tones is set to "0" (14-2). When the condition of 14-1 is not satisfied, as can be seen from 12-2, since the current position i is the timing of generation of a harmony, the next position (i + 1) is set as the search start position inext, and the function f ₁ indicating the situation in front is shown. Is set to “1” to indicate that it is a harmony, and the non-harmony number counter inrs
c is set to “0”, and the counter arpc of the number of harmony sounds in the bar is incremented. (14-3). Suppose "-1" indicates that the function f ₂ showing the rear situation is barline at 14-4, 14
At -5, j is set to the search start position.

Hereinafter, in the loops 14-6 to 14-10, j = inext to BEA
Move to T (see 14-9 and 14-10). In the meantime, when a non-harmonic sound generation timing is found, the counter rsc of the non-harmonic sound in the bar and the counter inrsc of the non-harmonic sound between the harmonic sounds are incremented. , The corresponding non-harmonic identifier RSrsc is set in INRSinrsc, and the position j at which the non-harmonic sound is found is temporarily stored in J (14-7, 14-8).
When the harmonic tone is found in the middle of moving to j = inext~BEAT rewrites the function f ₂ showing the rear of the status "1" (indicating a harmonic sound). Therefore, when no harmony is found until the end, f ₂ = −1.

At 14-2, the pitch MDarpc of the front harmony is set to bef, and at 14-3 the pitch MDarpc + 1 of the rear harmony is set to aft.

FIG. 15 is a flowchart for adding one non-harmonic sound. In this process, a pitch candidate of a non-harmonic tone is selected, a function f is calculated for this candidate, the previous inference is executed, and the analysis result is stored in the non-harmonic tone temporarily stored in the preprocessing 14-8. A match is considered successful if it matches the element INRS ₁ in the pattern. Non-harmonics are not added when all candidates fail. The pitches of the non-harmonic candidates are set based on the high temperature of the preceding and succeeding harmony. When j = 1, an attempt is made in the vicinity of the forward harmony, and when j = 2, an attempt is made in the vicinity of the rear harmony.

According to the flow of FIG. 15, j is first set to 1 (15-1). In 15-2, when j = 1 and moreover f ₁ is -1 indicating the measure line, or in 15-3 j =
2 f ₂ proceeds to the next processing when the -1 indicating the bar line (15-15,15-16). After setting k = 1 at 15-4,
In the loop from 15-5 to 15-14, move d as follows (see 15-5): k: 0 1 2 3 4 5 ... d: 0 1-1 2-2 3 ... where d is Represents the pitch difference between a non-harmonic candidate and a forward or backward harmonic. That is, when j = 1, the forward chord sound
The sound separated by d from bef is selected as the pitch md of the non-harmonic candidate, and when j = 2, the sound separated by d from the backward chord is selected as the pitch md of the non-harmonic candidate (15-6. 15-8).
Then, it is checked whether or not md is a scale sound and a non-harmonic sound in 15-9. Whether or not md is a non-harmonic tone is determined from the corresponding code CDi (CDi), and the code number at the current position is obtained by referring to the sequence {CRi} of the code length), and whether or not it is a scale sound Kaha Code CDi
Is determined from the scale data (not shown) corresponding to. Then the rest of the function f ₃ at 15-10, is calculated for f _4, f _5. The function f ₃ is the difference between md and bef, the function f ₄ is the difference between md and aft, and the function f ₅ is 0 when f ₃ × f ₄ is positive, and 1 when it is not. All functions representing the melody situation
Since f ₁ ~f ₅ is determined contrast, performing countercurrent inference before applying the classification knowledge nonharmonic tones determine the type of nonharmonic tones (15-11). Regarding this inference, Japanese Patent Application No. 62-325178
Since it is shown in the issue, only a brief explanation is given. FIG. 17 shows production rule data expressing non-harmonic classification knowledge, and the knowledge is represented by a tree in FIG. In forward-looking inference, start with Rule No. 1
Checked for establishment of preamble L ≦ f _x ≦ U rules indicated by the (f _X type of function values indicated by the rule data X is), the rule data N is not established if negative part positive portion rule data Y if satisfied If the rule data Y or N is a negative value, the processing is continued, and the inference is completed with the absolute value as the identifier (analysis result) of the non-harmonic tone. The analysis result and the non-harmonic tone identifier FNR obtained in the preprocessing
It is compared with S ₁ , that is, the type of non-harmonic sound to be generated at the focused non-harmonic sound generation timing, and if they match, it is determined to be successful (15-12). If unsuccessful, k is advanced to check for the next pitch candidate (15-4). If the failure at 15-12 continues for k = 1 to 10, j is incremented to 2 (15-13, 15-16). Even when j = 2, k is between 1 and 10.
If unsuccessful in 15-12 continues, the flow exits assuming that no non-harmonic sound is added. (15-15).

FIG. 16 is a flowchart when two non-harmonic sounds are added. In this example, the first md1 is set from the preceding pitch data bef, and the second
md2 is set from the pitch data aft that follows. (16-3 ~
16-6), j is a variable for the first pitch candidate,
k is a variable for the second pitch candidate. Both non-harmonic candidates md1 and md2 are non-harmonic pattern elements IRNS
(IRNS ₁ and IRNS ₂ ) are regarded as success (16-
10, 16-14, 16-10).

The other points are clear from the above description and the description in FIG. 16, and therefore the description is omitted.

<Data Format Conversion> FIG. 19 shows a flow of the data format conversion executed in 9-4 of melody generation (FIG. 9). At the stage of entering this flow, RRARP contains information on the timing of the generation of the harmony in the bar, {MD} contains the pitch data of the harmony generated at the timing of the generation of each harmony, and RRNCT contains the pitch data of the The non-harmonic sound generation timing information and the {MD} contain the pitch data of the non-harmonic sound generated at each non-harmonic sound generation timing. Thereby, the melody in one bar is completely exposed. In the processing of FIG. 19, data conversion is performed to form a pitch sequence {VMED} and a pitch sequence {VMER} of these information melodies.

In the figure, s is a counter for obtaining the duration of each sound, mdc is a counter for harmony, ndc is a counter for non-harmonic, vmedno
Is a melody sound counter, VMERi represents the pitch of the i-th melody sound, and VMEDi represents the pitch of the i-th melody sound.

At 19-1, s = 1, mdc = ndc = 0, vmedno = 0 are set, and at 19-2, i = 1 is set. Move i from 1 to BEAT in the loop from 19-3 to 19-8 (19-7, 19-
8) During that time, when a harmony is found (19-3: YES), the counter mdc of the harmony is incremented, the duration s of the previous tone is set to VMERvmedno, and vmedno
Pitch data MDmd of the harmony sound found as an increment
c is set to VMERvmedno, and the sound length counter s is returned to “0”. (19-4). When a non-harmonic tone is found (19-5: YES), the counter ndc of the non-harmonic tone is incremented, the duration s of the previous tone is set to VMERvmedno, and vmed
Non-harmonic pitch data found by incrementing no
Set NDndc to VMERvmedno and reset s to “0”. The value of s is returned to “0” when the occurrence of a harmony or non-harmony is detected, and is incremented at 19-7, so that the length from the detection of a sound to the detection of the next sound, that is, Indicates the length of the sound. When i> BEAT is satisfied in 19-8, the length s of the last sound is set to VMERvmedno (19-9), and the flow exits.

Frequency change As shown in the main flow (FIG. 3), when the user inputs a judgment result of “like” or “dislike” to the generated melody, the frequency in the frequency data storage device 30 is determined according to the judgment result. The data changes. FIG. 20 shows the flow of the frequency change processing 3-9 and 3-10.

20-1 to 20-3 are to analyze the melody of the determined bar, 20-1 is to analyze the harmony included in the melody, 20-2 is to analyze the non-harmonic included in the melody, In 20-3, non-harmonic sounds included in the melody are analyzed. The data obtained by the melody analysis, that is, the values of the melody control parameters used for generating the melody related to the determination define the object to be changed in the frequency data stored in the frequency data storage device 30. 20-4 shows that the frequency data to be changed is changed in accordance with the result of judgment after being input.

This data change in 20-4 is performed, for example, as follows. Now, suppose that the result of the melody analysis (20-1 to 20-3) is that the value of the control parameter of a certain type i (to be represented by Pi) is k. On the other hand, among the frequency data DATA _{i, j} (j = 1 to e) for the same type i in the frequency data storage device 30, the data DATA _i, k of j = _k is read out, and the judgment result is “like” ", The data DATA _{i, k} are incremented,
If the judgment result is "dislike", the data DATA _{i, k} are decremented. This processing is performed on the data indicated by the value of the parameter of the corresponding type obtained by the melody analysis, out of the frequency data for each type.

Therefore, as for the value of the control parameter which is the generation factor of the melody determined to be "like", the value of the corresponding frequency data becomes higher, and as a result, when a new melody is generated, this is the generation factor. As the value of the control parameter, a value that the user judges as “like” is likely to occur. On the other hand, regarding the value of the control parameter, which is the cause of the melody that the user has judged as “dislike”,
Since the value of the frequency data at which the value is generated becomes low, the value is less likely to be generated when the melody is subsequently generated. This means that as the operation continues, the possibility of generating a melody that suits the user's preference is increased, giving the composer a self-composing music ability that is highly desirable for the user. Is what it is.

Hereinafter, the harmony analysis 20-1 and the non-harmony analysis 20-2 and the rhythm analysis 20-3 in the frequency change will be described in detail.

<Harmonic Analysis> FIG. 21 shows a flow of the harmony analysis processing. The purpose is to obtain the pattern {LL} for the harmony contained in the determined melody.

In the first 21-1, the counter llno of the number of harmony sounds found in the bar is initialized to 0, and in 21-2, the counter i of the basic time length in the bar is initialized to 1 which is the head of the bar.

Then, it is checked at 21-3 whether a melody note exists at the i-th position in the measure to be evaluated.

This test is This is realized by checking the presence or absence of N that satisfies the condition. Here, BEAT is the number of basic unit times of one bar, and MRj is the duration data of the j-th melody of the music. Two
When the condition of 1-3 is satisfied, the note number and chord number corresponding to the position (measure number × BEAT) +1 of interest at present are obtained in 21-4.

this is, M (same as N above) that satisfies This is performed by finding C that satisfies Here, {CRj} represents a sequence of chord lengths of chord progression. Since the code length of this progress is variable, the code CDc at the melody occurrence position is obtained by referring to the column information {CRj} of the code length.

Next, at 21-5, the melody tone MDm (the m-th tone) focusing on the constituent tones of this chord CDc (the root note and type data of the c-th chord, see FIGS. 4 and 5) (Ie, data representing the pitch), that is, whether the melody sound is a harmony sound. In detail, this processing exchanges the data format of the code CDi into the data format of the constituent sounds (data in which each digit of 12 bits represents a pitch name), converts the melody pitch data MDm into a similar data format, Is performed by checking whether the result takes a value other than 0. For example, the constituent sound data of the major chord of C is given by 10010001 and the sound of E is given by 10,000, and when it is ANDed, it becomes 10,000.
It can be seen that this is a constituent sound (harmonic sound) of the major chord.

If it is determined in step 21-5 that the harmony is a harmony, the harmony counter llno is incremented in step 21-6, and in step -7, the octave of the harmony indicated by MDm from the code CDc and the melody tone MDm. , The octave information of the LLllno is obtained from the upper 8 bits of the MDm, and is substituted for the octave information of the MDm. The number of the harmony to be located in is obtained by comparing the note name information in the lower 8 bits of MDm with the constituent sound information of the code CDc. For example, in the above example, the sound E is the second harmonic sound from the bottom (see LLi in FIG. 6).

Next, the position of interest is shifted to the next position in 21-8, and the processing of 21-3 and below is repeatedly executed from i = 1 to BEAT (21-9).

<Non-harmonic Analysis> FIG. 22 shows details of the non-harmonic analysis 20-2.

Steps 22-1 to 22-5 correspond to the above-described steps 21-1 to 21-5. Here, rsno 21-1 is a counter for the number of non-harmonic tones found in the bar. The purpose of the processing is the parameter {RS} of the non-harmonic sound included in the melody.
Therefore, when the sound of interest is determined to be a non-harmonic sound in the determination of 22-5, the process proceeds to 22-6 and below.

In 22-6, the value of the function representing the melody situation before and after the non-harmonic tone of interest is calculated. As described above, an example of this function is f ₁ : whether the preceding note is a harmony or not f ₂ : whether or not the next note is a harmony f ₃ : pitch with the preceding note Difference f ₄ : Pitch difference from the next note f ₅ : Whether the hour and minute is a pole (if f ₃ × f ₄ <0, pole).

Next, after incrementing the non-harmonic counter rsno in 22-7, the above-described non-harmonic classification knowledge (production rule data) is applied to the function result in 22-8, and the type of the non-harmonic is set to the previous value. Calculate by direction inference. This forward-looking inference
The identifier (number) of the non-harmonic sound, which is the inference result of No. 8, is assigned to the register RSrsno of the non-harmonic sound pattern element (22−
9). Then, the counter i of the basic time in the bar is incremented (22-10), and the processing of 22-3 or less is repeated until i = 1 to BEAT.

<Rhythm analysis> The flowchart of rhythm analysis 20-3 (Fig. 20) is shown in Fig. 23.
Shown in the figure. The purpose of this processing is to extract the generation timing information of the harmony sound, the generation timing information of the non-harmonic sound, and the generation timing information of the rest included in the melody.

Processing 23-1, 23-2, and 23-3 are harmony analysis 21-2,
Since they correspond to 21-3 and 21-4, the description is omitted.

If the note MDm of interest in 23-4 is a harmony, 2i ^-1 is added to RRARP in 23-5. If the note MDm is nonharmonic tones in 23-6 adds 2 ^i-1 to RRNCT Te 23-7.
When MDm is a rest (when 23-6 is not satisfied), 2i ^-1 is added to RRRST. This processing is performed for i = 1 to BEAT
Repeat (23-9, 23-10). As a result, RRART obtains the timing information of the generation of the harmonies included in the input melody, RRNCT obtains the generation timing information of the non-harmonic sounds included in the input melody, and RRRST obtains the generation timing information of the rest. can get. Each bit of RRRART, RRNCT, and RRRST represents each timing in the bar, and bit “1” indicates that a harmony, a non-harmony, and a rest are generated at that position.

Modifications Although the description of the embodiments has been finished above, the present invention is not limited to the above-described embodiments, and various modifications and changes are possible.

For example, the present applicant has proposed an automatic composer that calculates a melody control parameter for each section in order to provide a hierarchical structure of music and generates a melody based on the calculation result (Japanese Patent Application No. 62-121037). ). This automatic music function can be combined with the embodiment described above. In this case, a melody of a certain section is generated from the frequency data, and a melody of another section is generated with reference to a melody of another section already generated. For example, when generating a melody for 8 bars,
In the first and fifth measures, a melody is generated based on the melody control parameters obtained from the frequency data without referring to the melody of the other section. For the second to fourth measures, the melody (motif) of the first measure is compared with the melody (motif) of the first measure. In consideration of the relationship, the melody control parameter calculation means used for generating the melody of the first bar converts the melody control parameters for the second bar, the third bar, and the fourth bar, respectively, and performs the melody control for each bar. Generate a melody from the parameters,
Measures are generated in consideration of the relationship with the fifth measure.
In an actual operation, for example, if the user makes a decision on the generated melody from the first measure to the fourth measure, the frequency data is updated. Again the first
When a composition instruction is given from a measure, the melody of the first measure is generated based on the updated frequency data, and the melody control parameters used for the generation are converted into information for the second, third, and fourth measures. Thus, the melody of the second, third, and fourth measures is re-created.

According to this combination, it is possible to generate a melody that suits the user's taste while giving the music variety and consistency, which is more desirable.

In addition, the present applicant has proposed an automatic composer having means for generating a melody control parameter for each section based on an analysis result of chord progression (Japanese Patent Application No. 62-325176). The function of this composer is also described above. Can be combined with the above embodiment in the same manner as described above. Further, the applicant of the present invention can control the value of the melody control parameter from the user side by enabling the weight of the control parameter element calculated by the melody control parameter calculation means for each section to be specified by an input from the user. An automatic composer is proposed (Japanese Patent Application No. 63-12476), but it can be combined with the above embodiment in the same manner.

Further, the applicant has proposed an automatic composer that receives a melody that forms part of a song from a user, analyzes the input melody, extracts melody control parameters, and generates a melody based on the extracted result. Yes
No. 62-86571), the present invention is also applicable to this type of automatic composer. In this case, if the melody analysis means 7 used for changing the frequency data in the above embodiment is also used for analyzing the input melody, the resources can be effectively used.

However, the melody analysis means 7 for changing the frequency data
Is not always necessary, and when the melody synthesis control parameter generating means 1 operates during melody synthesis, its generation result is stored for each section, and when the judgment result for the melody of a certain section is given from the input means 6, The melody control parameter to be read may be read.

Also, the unit of the melody length determined by the user may be the same unit as the melody generation unit as in the above embodiment, or may be the unit of another section (for example, a passage). .

Further, the form of input of the user's determination result may be direct or indirect. To give an indirect example, when the user instructs to generate a melody, when the user instructs to generate the same section as before, the user's judgment on the melody in that section was negative (disliked). It is considered that when the generation instruction of another section is given, the user's judgment on the melody of the section generated before that is good (like). Further, when a melody of a part of the music is input from the user, the melody in that section can be processed as if it is always a favorite judgment.

Further, for a specific section (for example, a rust section), frequency data different from frequency data that is basic information for generating a melody in a normal section may be used.

[Effect of the Invention] The automatic music composer according to the present invention "learns" the user's preference for the melody once generated, accumulates the information in the information called frequency data, reflecting the preference, and generates a new melody. At this time, a melody control parameter is generated based on the frequency data, and a melody is generated based on the generated melody control parameter. Therefore, songs that suit individual users' preferences,
Alternatively, it is possible to efficiently compose a song that matches the user's idea of each song, and achieve an important ability required for a composer.

[Brief description of the drawings]

FIG. 1 is a functional block diagram of an automatic composer according to an embodiment of the present invention, FIG. 2 is an overall configuration diagram of the automatic composer, and FIG.
Fig. 4 is a flowchart showing the overall operation of the embodiment, Fig. 4 is a diagram showing a variable list, Fig. 5 is a diagram showing a partial variable data format, and Fig. 6 is a diagram showing another partial variable data format. ,
FIG. 7 is a diagram showing a storage format of frequency data stored in the frequency data storage device, FIG. 8 is a flowchart of reading frequency data, FIG. 9 is a flowchart of melody generation, and FIG. 10 is a flowchart of harmony generation. , Eleventh
Fig. 12 is a flowchart of data conversion from LL to MD. Fig. 12 is a flowchart of non-harmonic sound addition. Fig. 13 is correspondence between harmony sound generation timing information RRARP and pitch data MD and non-harmonic sound data generation timing information. FIG. 14 is a diagram showing correspondence between RRNCT and pitch data ND, FIG. 14 is a flowchart of preprocessing, FIG.
FIG. 15 is a flowchart for adding one non-harmonic sound, FIG. 16 is a flowchart for adding two non-harmonic sounds, FIG. 17 shows an example of production rule data expressing non-harmonic classification knowledge, and FIG. FIG. 17 shows the meaning of the production rule data shown in FIG. 17, FIG. 19 is a flowchart for converting into pitch sequence and pitch sequence data, FIG. 20 is a flowchart for frequency change, FIG. 21 is a flowchart for harmony analysis, 22nd
Fig. 23 is a flowchart of non-harmonic analysis, and Fig. 23 is a flowchart of rhythm analysis. 1 ... melody generation control parameter generation means, 2 ... chord progress storage means, 3 ... melody synthesis means, 5 ... monitoring means, 6 ... input means, 7 ... melody analysis means, 8 ... data change means , 9 ... frequency data storage means, 10 ... CUP, 20 ... frequency data storage device, 40 ... input device, 50 ... monitor.

Continuation of the front page (56) References JP-A-62-187876 (JP, A) JP-A-60-107079 (JP, A) Takasawa, Yoshikawa "Automatic performance and composition of electronic organs by computer", Computepia stock Computer Company, December 1, 1975, December 1975, Vol. 9, No. 110, p. 13-19 (58) Field surveyed (Int.Cl. ⁶ , DB name) G10H 1/00 102 G10H 1/36

Claims (3)

(57) [Claims] 1. A melody control parameter generating means for generating a plurality of types of melody control parameters, and a melody synthesizing means for synthesizing a melody based on a melody control parameter generated by the melody control parameter and a given chord progression. In the automatic composer provided, frequency data storage means for storing frequency data for controlling the frequency at which various melody control parameters occur at specific values by type and value, and a melody synthesized by the melody synthesis means. Output means for outputting; input means for inputting a user's judgment result on the melody output by the output means; and melody output by the output means among frequency data stored in the storage means. Corresponding to the values of the various melody control parameters used for Frequency data changing means for changing the radiated frequency data in accordance with the judgment result input by the input means, wherein the melody control parameter generating means newly generates the melody control parameter based on the frequency data changed by the frequency data changing means. An automatic composer characterized by generating a melody control parameter for controlling a melody to be generated. 2. A melody control parameter generating means for generating a plurality of types of melody control parameters, and a melody synthesizing means for synthesizing a melody based on a melody control parameter generated by the melody control parameter and a given chord progression. In the automatic composer provided, frequency data storage means for storing frequency data for controlling the frequency at which various melody control parameters occur at specific values by type and value, and a melody synthesized by the melody synthesis means. Output means for outputting; input means for inputting a user's judgment result on the melody output by the output means; and melody output by the output means among frequency data stored in the storage means. Corresponding to the values of the various melody control parameters used for Frequency data changing means for changing the radiated frequency data in accordance with the judgment result input by the input means, wherein the melody control parameter generating means newly generates the melody control parameter based on the frequency data changed by the frequency data changing means. A melody control means for generating a melody control parameter for controlling a melody to be generated, wherein the melody synthesizing means synthesizes the melody with the generated melody control parameter; Melody input means for inputting a melody; and input melody analysis means for analyzing a melody input by the melody input means and extracting an input melody feature parameter. Melody that does not generate melody control parameters based on For while an automatic composer which is characterized in that it comprises means for generating a melody control parameters based on the input melody feature parameters extracted in the input melody analysis means. 3. A melody control parameter generating means for generating a plurality of types of melody control parameters, and a melody synthesizing means for synthesizing a melody based on the melody control parameters generated by the melody control parameters and a given chord progression. In the automatic composer provided, frequency data storage means for storing frequency data for controlling the frequency at which various melody control parameters occur at specific values by type and value, and a melody synthesized by the melody synthesis means. Output means for outputting; input means for inputting a user's judgment result on the melody output by the output means; and melody output by the output means among frequency data stored in the storage means. Corresponding to the values of the various melody control parameters used for Frequency data changing means for changing the radiated frequency data in accordance with the judgment result input by the input means, wherein the melody control parameter generating means newly generates the melody control parameter based on the frequency data changed by the frequency data changing means. A melody control parameter for controlling a melody to be generated is generated, and the melody synthesizing means is an automatic music machine characterized by synthesizing a melody with the generated melody control parameter. An automatic composer comprising melody analysis means for analyzing the output melody in order to obtain a melody control parameter used for synthesizing the melody output by the output means.