JP2666063B2 - Automatic composer - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Technical Field of the Invention]   The present invention relates to an automatic composer. [Background]   The sound that makes up the melody line of normal music is a harmony
And non-harmonic sounds. Focusing on this point,
The present applicant filed a Japanese Patent Application No. 62-86571 (April 8, 1987)
Application), given a motif
The characteristic parameters of the dispersed chord included in the motif
Extract the characteristic parameters of the harmony and follow the extraction result.
In each melody generation section,
To generate a dispersive chord, and then add a non-chord
To apply for an automatic composer that completes the melody sequence
I have.   Therefore, this automatic composer can play harmony-friendly songs.
It is possible to compose. On the other hand, each melody generation section
The code for
Must be selected from the list. This means
This will add harmony restrictions to the generated melody.
You. Also, in order for the user to enter the chord progression,
Requires knowledge of harmony as background. [Object of the invention]   The present invention has been made in view of the above background, and has been developed.
The goal is to range from harmonious to non-harmonic
It is possible to compose music widely and to use special music knowledge
It is to provide an automatic composer that is unnecessary. [Structure and operation of the invention]   According to the present invention, the mochi
Motif input means for inputting music and one for each music section
Or the primary constituent sound consisting of multiple note names
Primary structure given as the primary component sound progression over the music section
Means for giving a sound progression, the motif and the primary constituent onset
A predetermined part of the line, corresponding to the time length of the motif
Motif and the primary constituent sound
The first characteristic parameter that shows how is distributed
Sound in the motif other than the primary constituent sound
And a second feature parameter indicating whether the distribution is
Characteristic parameter generating means for generating the first characteristic;
Change the parameters by random numbers,
A first stage consisting of only primary constituent sounds based on the primary constituent sounds
First stage melody generating means for generating a melody of the floor;
The melody of the first stage according to the second feature parameter
A sound other than the primary constituent sound is added to the
Melody generating means for outputting the melody
Is provided.   This automatic composer performs melody generation in two stages.
The melody of the first stage is the final melody
The melody of the second stage is equipped with this skeleton.
It works as a decoration or a connection. This double melody
Line, what sound to choose as the primary constituent sound
Change, and in some cases, traditional songs
In some cases, new and experimental songs are formed. [Example]   Hereinafter, examples of the present invention will be described. Composer Mo
In this mode, the automatic composition device converts the sound to the first stage melody sound.
And the second stage melody sound
Has adopted. The first stage of the composition
The primary constituent sound that determines the pitch organization of the scale melody sound (hereinafter
Progress information on motifs (called sounds)
Melody), the rhythm or length sequence of the melody to be generated
Scale and basic scale type used to control the sound
Is given. Each sound included in the motif is divided into sections.
By using the applied double tone information, the first stage melody sound
The second stage melody sound is identified. Second from motif
The part excluding the step melody sound is the first step melody of the motif.
This is the sound sequence of the D sound. From this sequence, the first stage melody
Are extracted. Also, the first stage melody sound
And the motif that distinguishes it from the second stage melody sound
On the other hand, the music knowledge (classification
Is stored in the production rule data memory described later.
Is included in the motif by using
Can extract the type (characteristic) of each second stage melody sound
You. That is, what kind of second stage is in the motif
Information indicating how the melody sounds are distributed (No.
Characteristic of a two-step melody). More double notes
Extract the hierarchical structure and tonality structure of the music.   Melody generation consists of the first stage melody generation process,
It consists of a two-step melody addition process and a tone sequence generation process.
You. In the process of generating the first melody,
According to the hierarchical structure extracted from the above, the first stage melody
The generation of the sound pattern is controlled. New first stage due to hierarchical structure
When the generation of the sound pattern of the floor melody is instructed, the first stage
Features of Lodi's sound pattern (first stage extracted from motif
From the characteristics of the Rody's sound pattern or its modifications)
One-step melody sound pattern is generated and the generated first step
By applying double tone information corresponding to the melody type
A first stage melody represented by a pitch sequence is generated.
The first stage melody generated in this way
A two-step melody is added. Add second stage melody
To do so, the music knowledge described above is used again. Second stage
In the inference of melody sound addition, the sound that can be added is the second stage
It must satisfy the characteristics of the scale melody, and
There is a condition. The scale note here is a double
Tonic of the basic scale according to the tonality structure extracted from the line
(Key, tonality) is the sound on the scale that is inverted. Second stage
Classification of melody sounds and addition of second stage melody sounds and smell
Because common music knowledge is used for inference,
"Reversibility" between melody analysis and synthesis
Given. Complete reversibility is the analysis of a melody
As a result, a certain analysis result was obtained.
To generate a melody,
Is consistent with the original melody. First stage melody
Melody by adding a second stage melody
Is completed. On the other hand, the melody pitch sequence
It is obtained as follows. That is, the basic rhythm
(Length sequence) is set to the target number of notes (for example,
Total number), use a pulse scale to optimize
Or optimal split. And where the note is split or
Is connected to each pulse of the selected pulse scale
It depends on the weight of the point. Therefore a consistent rhythm
Control is possible. <Overall configuration>   FIG. 1 shows an overall configuration diagram of the automatic music composer according to the present embodiment.
You. CPU1 is used to implement the automatic music function in this device.
Control device. From the input device 2, a motif (melody)
Di), compound sound progression, type of pulse scale used, usage
The type of scale to be used is input. Motif memory 3
Stores the input motif. Polyphonic memory
Reference numeral 4 denotes a memory for storing information on the progression of multiple sounds.
The information is extracted the first stage melody when analyzing the polyphonic progression.
Used by CPU 1 when issuing or generating. sound
The scale data memory 5 is a scale representing various kinds of scales.
This is a memory for storing data.
Sets a specific scale as a scale to be used in a song.
It can be selected from a scale set. Product
2nd stage melody sound
Music knowledge for classifying is stored. This music knowledge
Is a place to classify the second stage melody sounds included in the motif
If the second stage melody is added to the generated first stage melody
Used when adding. The pulse scale memory 7
Remembers the pulse scale (set of pulse scales)
This is a memory that allows the user to place the song at the beginning of composition.
Considering the characteristics of the rhythm,
You can select the desired pulse scale from
Wear. The selected pulse scale is the melody rhythm
(Length sequence). Melody data memory
8 stores the completed melody data. External memory
The melody data stored in the melody data memory 8
Transcripts, different musical knowledge, different composition program resources and
To be used. Work memory 10 is used while CPU 1 is operating.
Various data to be used, such as tonal structure, hierarchical structure, various
Variables and the like are stored. Monitor 11 is CRT12, staff notation
It consists of a linter 13, a tone generation circuit 14, and a sound system 15.
The results of composition and analysis are displayed through these devices.
Can be displayed and output. <Composition General Flow>   Overall flow of this device in the composer mode (composition general
Ralflow) is shown in FIG.   In the initial setting in 4-1
Then, basic information for composition is input. The user
As information to inform the automatic composer, (1) BEAT, (2)
Type of pulse scale (3) Scale (4) Fully automatic or mochi
Input is shown. BEAT is a note of the basic unit length (the shortest
Note) is the length of one measure expressed by the number of
This defines the time signature of the song. For example, 4 time system
For the song of BE, the basic unit length of the note is the 16th note BE
If AT = 16 is set, the length of one bar is 4 beats. 4-1
Select the pulse scale of the song composed by the automatic composer.
This is information that primarily controls the rhythm. Pulse scale
Is a combination of notes at each pulse point with the interval of the basic unit length.
Weight, which indicates the ease of dividing or dividing notes
This is the scale attached (see Figs. 7 and 12).
The melody duration by using the pulse scale
Columns are controlled. Therefore, the type of pulse scale
Choosing is the rhythm of the song that the automatic composer composes.
This is equivalent to selecting the feature of the program. Select with 4-1
Type of scale to be played (eg diatonic scale)
Is the scale used by the automatic composer in composing music. Sa
In the initial setting 4-1, the composition is performed automatically.
It is determined by the user whether to perform the operation based on the motif.
When fully automatic, (1) multiple data
Sound progression, (2) Production rules (3) Pulse schedule
Is read into the work memory 10 (4-3),
(1) Basic rhythm (length pattern), (2) First
Features of the step melody pattern (PCi: see Fig. 5),
(3) From the characteristics of the second stage melody (RSi: see Fig. 5)
Is generated according to the user's instructions (4
-4). On the other hand, in composition mode using motifs,
Motif other than the above data as data reading 4-5
(Input melody) is also read, as an essence
(1) Rhythm (2) First stage melody pattern (3) No.
Features of one-step melody pattern, (4) Second-step melody
The features of the motif are extracted from this motif (4-5).
In both automatic and motif use modes, double-tone
A row evaluation 4-7 is performed, where (1) hierarchical structure (2)
The tonality structure is extracted from the polyphonic progress information. This hierarchical structure
Expresses the coherence and diversity of the songs inherent in polyphonic progression
It is. In addition, the tonality structure indicates the sound used in each section.
It defines the key of the floor (fundamental tone). 4-7
Has completed the "analytical work" for composing
You. For example, the characteristics of the first stage melody pattern
The information required to generate the sound pattern of the step melody
The feature of the floor melody is the second melody added to the first stage melody.
It characterizes the step melody. Also tonal structure
Restricts the melody sound candidates in each section. Hierarchical structure
Determines whether to generate a new first-stage melody pattern
Can be used to decide. Also, the pulse scale
Used for generation. In melody generation 4-8, (1)
Generation of one-step melody (2) Addition of second-step melody
(3) Rhythm generation is performed. <Variable list, data format>   List of main variables used in the flow described later
The data format is shown in FIG. 4 to FIG. Data
The format is merely illustrative, and any other suitable data
Format is available.   Hereinafter, the composer mode of the present apparatus will be described in detail.
You. <Initial setting>   Initial setting 4- in the composition general flow (Fig. 2)
9 is illustrated in detail in FIG. Select with this initial setting
BEAT meaning, pulse scale type (PULS), scale type
(ISCALE) meaning, fully automatic or motif input meaning
Since it has been described with reference to FIG. 2, the description is omitted here. PULS
The values are stored in the pulse scale memory (Fig. 1).
Pointer to a certain type of pulse scale, ISCALE
Is a specific type of sound stored in the scale data memory 5.
Pointer to floor data. <Read data>   Initial setting as shown in the composition general flow of Fig. 2.
After that, data reading is executed in 4-3 or 4-5.
Is performed. In the case of fully automatic, the
Since the chief is not used, reading the motif data
Not done. The reading of each data is explained below.
You.   FIG. 10 shows the multiple sounds stored in the multiple sound progression memory 4 (FIG. 1).
An example of sound progress data is shown. Fig. 11 shows the note progression memory 4.
In the flowchart to load the remembered double tone progress data
is there. In the memory map shown in FIG.
The sound data (CCi) is placed at an even address and its double tone
Is located at the next address (odd address)
You. For example, a CCi with a value of 8A4 in hexadecimal notation is G, B,
CRi, which represents four sounds D and F, and has a value of 10,
The length is 16 times the basic time length (for example, 16th note)
Is represented.   In FIG. 11, the register CCi has the i-th
The double tone data that appears in the eyes is entered, and the length is entered in CRi.
You. CDNO contains the total number of multiple sounds. About other points
It is clear from the description in FIG. 11 that the description is omitted.
You.   FIG. 12 shows the data stored in the pulse scale memory 7 (FIG. 1).
An example of pulse scale data is shown below. Fig. 13 shows the pulse
Of the type selected in the initial setting from the scale memory 7
It is a flowchart which loads a loose scale. this
In the case of the example, the rhythm characteristics of the song to be composed by default
Select the pulse scale type (PUL
S) is a specific address in the pulse scale memory 7.
(For example, 0), and the address
Contains the start address of the pulse scale data
You. Configure a pulse scale for this start address
Sub-scale (scale with only 0 and 1 weights)
The number is stored, and the data of each subscale is stored in the subsequent address.
Is stored. For example, normal pulse scale
Is the five subscales FFFF, 5555, 1111, 0101, 0001
(Hexadecimal representation) and the corresponding binary representation is shown in Figure 7.
I have. For a normal pulse scale, the first pulse
The weight of the loose point (the rightmost position in FIG. 7) is 5 which is the maximum.
This means that the normal pulse scale
If you select this, the maximum of each section (measure) of the generated rhythm
This indicates that the note is most likely to exist in the first position.
You.   FIG. 14 shows the production rule data memory 6 (first
Figure) shows an example of production rule data stored in
doing. FIG. 15 shows the data stored in the memory 6.
It is a flowchart which reads. Production rules
Of the second stage melody sound included in the melody
It expresses musical knowledge to be similar to each product.
Action rule data is data that defines the prerequisites for the rules.
Function data indicating the lower limit data Li and the function type as data
Data Xi and upper limit data Ui.
Data Yi and Ni. Function is the shape of the melody to be analyzed
This is a numerical representation of the situation, an example of which is shown in Figure 29 below.
Is done. Whether the value Fxi of the function indicated by the data Xi is greater than or equal to Li
Is less than or equal to Ui (Li ≦ Fxi ≦ Ui).
Is a prerequisite (proposition) of the rule
The conclusions are shown in data Yi,
The conclusion when standing is shown by data Ni. And de
If the data Yi or Ni has a positive value,
Number of the next production rule to refer to
And when it has a negative value, the absolute value
The type of the step melody sound is expressed. Forward inference must be 1
Start with one rule, and refer to this rule as the root
Call. Before finding the negative conclusion Yi or Ni
Directional inference ends.   The address of the production rule data shown in FIG.
For quotas, the data for each production rule is
5 consecutive addresses starting from the address of data Li
Is stored. For details, Li is an address divisible by 5.
Xi at the address of remainder 1 divided by 5, and the address of remainder 2
Ui to the address, Yi to the surplus 3 address, and the address to the surplus 4
The data of Ni is stored in the memory.   In FIG. 15, RULENO contains the production rules.
The total number is set. Other points are described above.
It is clear from the description of the flow.   Fig. 16 shows the motif data stored in the motif memory 3.
An example of data (melody data) is shown below. Figure 17 is based on the composition
It is a flowchart which reads the motif data.
In the case of FIG. 16, note pitch data MDi is stored at even addresses.
At the next odd address, the note length data MRi of the note is
It is remembered. The MDNO in Fig. 17 contains the number of notes in the motif.
Set.   This concludes the description of data reading. <Generation of essence>   Data in fully automatic composition mode without motifs
After reading, rhythm as the essence of the song, the first stage
Generate the characteristics of the melody pattern and the characteristics of the second stage melody
(4-4 in FIG. 2). Figure 18 shows this essence student
4 shows a flowchart of the generation. These essences are used
Generated by user specification or fully automatic. For example, 22
The setting of the reference rhythm pattern at -1
Pattern generation means, for example, 4/4 time signature, pulse
When Normal is selected as the scale Automatically generate a reference rhythm pattern.
Or, the user can enter a favorite rhythm pattern
Done. 22-2 First stage melody pattern features
And the feature setting of the second stage melody of 22-3
This is performed by an action or input by a user. Figure 19 shows random numbers
Automatically sets the characteristics of the first stage melody pattern due to occurrence etc.
FIG. 20 shows the flowchart for the second stage melody.
Float that is set by the user inputting the features of
An example is illustrated.   In the first stage melody pattern feature setting (Fig. 19)
PC₁~ PC_FiveIs the first row in a section of a predetermined length (eg, bar)
Number of notes that make up the floor melody, highest note, lowest note, and neighbors
First stage The maximum value of the difference between the melody sounds, adjacent first stages
Represents the minimum value of the difference between the scale melody sounds (Fig. 5
reference). Each PC can be generated for each section, and
Set the upper and lower limits of the value, and generate random numbers between them.
And obtained. Or you can use a PC sequence
Prepare the desired PC series in the database
You may do so.   In the feature setting of the second stage melody in Fig. 20, the monitor
With the key corresponding to the type a of each second stage melody sound
A word is displayed to prompt the user for input (24-
2). Second stage melody input by user in RSi array
The sequence of the sound identifier a is entered (24-3, 24-6, 24-7).
When the code EOI indicating the end of the input is called, a second
Enter the number of step melody sounds and exit the flow (24-
8). <Extraction of essence>   In the composition mode using motifs,
Then, from the motif, the essence of the song (rhythm,
Rody pattern, first stage melody pattern features,
The characteristic of the two-stage melody sound is extracted (FIG. 2, 4-
6).   Fig. 21 shows the rhythm evaluation of the motif, the first stage melody.
Fig. 25 shows the pattern extraction, the first stage melody pattern
Fig. 26 shows the feature extraction of (sound type) and the features of the second stage melody.
Fig. 27 shows an example of the feature extraction. In these flows, each
Is extracted for each section (measure).   In the rhythm evaluation of FIG. 21, Ps at 25-1
The number of the note at the beginning of the target bar
Position information, and Pss is the beginning of the measure indicated by Ps
Length of note protruding into the previous measure (basic time expression)
Is entered, and Pe is the last note of the target bar (the beginning of the next bar).
The position information of the note (one note before the note). 25-2
The indicated rr is a 16-bit value that stores the rhythm pattern of the target bar.
If the length of one measure is 16, the register
The first bit position of the star rr is the first basic time of the measure.
Similarly, the Nth bit position is Nth from the beginning of the measure.
Represents the base time of the eye. The processing from 25-3 to 25-9 is
The position of the note between the note Ps and the note Pe in the motif is
Determined using chief length data MRi and correspond to register rr
This is the process of filling in the bit position to be executed. For example, rr
hand, 0001000100010001 Is obtained, the pattern rr is
Sounds are generated at the first beat, second beat, third beat, and fourth beat of the knot.
And   Details of the calculation of Ps, Pss, Pe, and Pee are shown in FIGS. 22 to 24.
Is done. Pee is the next note after Pe, ie the beginning of the next bar
Indicates the length of the note that interrupts the target bar.   In the calculation flow of Ps and Pss in FIG. 23, beat is one small
Bar length (basic time length expression), bar is the number of the target bar
(Indicates the bar number specified by the user. The specified bar
Number is smaller than 1, but larger than the number of measures (mno)
Is an input error. If the specified measure is 1,
Ps = 1 and Pss = 0 are set (27-4, 27-5). Ps = 1
The reason is that in the case of the first bar, the first note of the bar
First note of the first note or the first note of all motif data
The reason for Pss = 0 is that the leading small
There are no clauses. A found in 27-2₁Is the head of the song
To the front bar line of the target bar, and this length
A₁Is obtained by accumulating the motif duration data MRi from the beginning.
(27-7, 27-8, 27-10, 27-1)
2). S = a₁Holds, the last accumulation in S
The next note after the i-th pitch data starts from the beginning of the target bar.
Start. Therefore, Ps = i + 1 and Pss = 0 (27−
9). On the other hand, S> a₁Is satisfied, S is added last.
Note with the note length data, i.e., the i-th note
This is the first note of the elephant bar. Therefore, Ps = i
You. Also, Pss = MRi-S + a₁(27-9).   The calculation flow of Pe and Pee shown in FIG. 24 is very similar to FIG.
Perform the following processing. Where a₁Contains the target bar from the beginning of the song
Explain the other points, including the length to the back bar line.
Description is omitted.   In the sound extraction process of the first stage melody shown in FIG.
The first stage melody pattern from the motif of the target bar
｛LLi｝ is extracted. To summarize the processing, Ps and P
For the measure indicated by e,
Using the corresponding double tone information, the motif data is
Determine whether it is a floor melody sound, and
For the determined sound, what is it from the bottom
Check if it is the second sound and get the data in LL format.
To be specific, first of all, from the motif data,
Find the first note (Ps) and last note (Pe) of interest
(29-1). Next, double tone data is loaded (29-
2). Next, the note counter i and the first stage melody sound cow
Then, the center k is initialized (29-3, 29-4). 29-5 processing
Is the pitch data MDi of the motif
This is a process for converting to data formation. For example, the sound "G"
It is converted into data mm having "1" at the bit position 7.   Check at 29-6 whether this pitch data mm is a double tone
I have a lock. This is the pitch data mm and the double tone data cc.
Can be determined by taking the logical product (mm∧cc) of 29-7 ~
In 29-13, in the bit “1” of the double tone data cc,
What number matches the bit “1” of the pitch data mm
Check if there is, and as a result c octave of the motif sound
Add the number (MDifooffoo) and add the first stage melody pattern
Data LLk. 29-15 counter to next note
Advance i, and change LL until the note number reaches Pe (29-16).
Ask. The first stage melody of the target section is
Enter the number of notes.   The characteristics of the sound pattern of the first stage melody in FIG.
First stage melody sound type extraction result {LLi}, from LLNO
Is derived.   In the feature extraction of the second stage melody in FIG.
Of the type of the second stage melody sound distributed in the motif between
Seeking a turn. Second stage melody sound counter and
Set the note counter (33-2, 33-3)
If the note is a second-stage melody note (33-4),
Calculate the function F that represents the situation of a motif centered on a note
And perform forward inference with production rules
Find the type of two-step melody sound and substitute it for RSj (33-5
~ 33-8). This second stage melody sound classification processing is Pe
By doing so, the array {RSj} contains the target
The second stage of the section motif
Is set. The 33--11 PSNO contains the second stage
Enter the total number of rody sounds.   Determine whether MDi shown in 33-4 is the second stage melody sound
Details of the separate processing are shown in FIG. This determination process is
Focusing on the extraction of one-step melody tone patterns (Fig. 25)
Whether the note being played is the first-stage melody sound (compound sound of multiple sounds)
This is the same as the process for determining whether or not
Whether the note name of the note of interest is included in the constituent notes
Can be determined by   In the calculation of the function F in 33-6, the forward inference
The mochi needed to classify the second stage melody sound
Calculate the conditions or factors of the melody. Fig. 29
FIG. 37 shows a specific example thereof. In this example, the function F
hand, F₁: How many notes ahead of the note (second melody sound) of interest
Is the first-stage melody sound (the first-stage melody
Di sound position) F_Two: First stage melody sound before the note of interest
Is located (the position of the first stage melody sound in front) F_Three: From the first stage melody sound to the first stage back melody
Number of second stage melody sounds before the sound F_Four: Forward first stage melody sound and backward first stage melody sound
Pitch difference with F_Five: Forward first stage melody sound and backward first stage melody sound
Of the pitch of the second stage melody between F₆: From the first stage melody sound to the first stage back melody
Whether the pitch of the melody up to the sound changes monotonically F₇: Pitch of the first rear melody sound and the previous sound
Difference (Pitch progression to the first stage melody sound backward) F₈: Pitch between the first stage melody sound and the one after it
Difference (Pitch advance from the first stage melody sound ahead) Is calculated (Fig. 29). Besides this, whether it ’s a weak beat or a strong beat
Information and information for distinguishing the duration are added to the set of functions F.
You may. The flowchart for calculating each function F
The description is omitted because it is clear from the description of itself.   Details of the forward inference in 33-7 are shown in FIG. Ma
First, move the rule number pointer P to the production
"1" indicating the rule that is the root in the
Set (44-1). After a while, the rule number poi
Premise (Lp ≦ Fxp ≦ Up) of the rule indicated by the counter P
Check if the rule is satisfied
Use the positive link data Yp as a pointer to the next rule
If the rule is not satisfied, the data of the negative conclusion
Use the data Np as a pointer to the next rule. However
When the data Yp and Np are negative values, the final conclusion is reached.
The absolute values (−Yp, −Np) in the second stage melody
Set to the conclusion register as an identifier of the sound type
You. According to the flow, the condition Lp> Fxp of 44-3 or the condition of 44-5
When the condition Fxp> Up is satisfied, the premise Lp ≦ Fxp of the rule P
≤Up is not established, so the data of the negative conclusion
Substitute Np into a (44-4, 44-6), otherwise
Is a prerequisite, so a contains the data of the positive conclusion of rule P.
Data Yp is entered (44-2). This a is substituted into P (44−
7) If P is a positive value, the next rule number pointer
Returns to the inspection of the next rule, and when P is negative, −P
Is the classification result of the second stage melody sound (44-8, 44-
9).   As an example of the classification of the second stage melody sound, the function F
In the calculation of F₁= 1: Back 1st stage Melody sound is the 2nd stage where attention is focused
One after the melody sound F_Two= -1: The first stage melody sound in front is the second stage of interest
One before the floor melody sound F_Three= 1: The number of notes between the first and second stage melody sounds is one F_Four= 8: The pitch difference between the first and second melody sounds before and after is 8. F_Five= 2: Second stage melody between first and second stage melody sounds
D sound is distributed in the middle of the pitch of the first and second melody sounds
doing F₆= 1: From the first stage melody sound to the first stage melody
The melody up to the tone has a monotonous change in pitch F₇= 1: It progresses by 1 pitch difference to the 1st backward melody sound
To F₈= 7: Proceeds with a pitch difference of 7 from the first stage melody sound ahead
Do Is obtained, and the production rule data
Let's make an inference using the data shown in Figure 14.   First, when P = 1 (route), the premise 0 ≦ F_Two≤0 Is F_Two= −1, it is not established. Therefore, le
Negative conclusion N₁= 3 is the point of the rule to be inspected next
Become   For P = 3, premise of rule 3 0 ≦ F₁≤0 Is F₁= 1, it is not established. Therefore, N_Three=
5 becomes the pointer of the next rule.   At P = 5, premise of rule 5 0 ≦ F_Four≤0 Is F_Four= 8, so it does not hold. Therefore, N_Five= 6
Becomes P.   At P = 6, premise of rule 6 1 ≦ F₆≦ 1 Is F₆= 1 holds. Therefore, Rule 6
Positive conclusion data Y₆= 7 is the next rule to access
Pointer P.   At P = 7, the premise of Rule 7 3 ≦ F₈≤∞ Is F₈= 7 holds. Where Y₇= -2 (negative)
is there. Therefore, conclusion = 2 (classified second stage melody
De sound identifier).   As you can see from this example, any type of second stage menu
Lody sounds have a finite number of propositions (the premise does not hold)
In other words, it means that the premise is false
) Can be identified by musical knowledge. This knowledge
The function F is calculated to represent
Rules have been created. That is, the function F is expressed in the second stage
The method used in the knowledge to classify the melody
Lodi check item information, production
Rule data leads to the classification result of each second stage melody sound
The sequence of rules up to is linked by a pointer. <Evaluation of double tone progression>   The device of this embodiment maximizes the progression of multiple sounds for composition.
Is one of the features. That is, FIG.
In the evaluation of the progression of multiple sounds shown in 4-7
And the hierarchical structure of the song
I want a structure and tonality structure. Hierarchy is consistent and versatile
The melody generation described later.
It is used for controlling the generation of the sound pattern of the step melody. On the other hand,
The gender structure represents the change in tonality (key) as the song progresses
Are used in each section in melody generation described later.
Used to select the scale key.   Hereinafter, the extraction of the hierarchical structure will be described in detail with reference to FIGS. 39 to 41.
Give details.   The flow shown in Fig. 39 is a block with a length such as a passage.
Calculates the similarity of the progression of multiple sounds between blocks using
It is a flow to be performed.   First, the song length SUM is calculated as the length of each compound sound CRi in the compound sound progress information.
Is obtained by accumulating (45-1), and barno (number of measures)
The length of the indicated block is converted to the
The lock length is set to 1 (45-2), and the music length SUM is set to the block length l.
To calculate the number m of blocks included in the song (45−
3). Counter i for reference number of block to compare
Is initialized to “0” (45-4).   In 45-8 to 45-18, the i-th block and j-th block
Calculates the similarity Vij of the progression of multiple sounds with the block (j ≧ i)
doing. As a function of similarity, You are using Where l is the length of the block, Vs
Is the compound of the i-th block and the j-th
Polyphony match obtained when comparing with block polyphony
Represents a number. This similarity function Vij takes a value from 0 to 100
So, at 100, the polyphonic progression of the two blocks is completely
(100%) match, and 0 means completely mismatch.   Calculation of similarity between i-th block and j-th block
Is started from the position of j = i (45-6), and the similarity is obtained.
Each time, the similarity with the next block is determined by j = j + 1
(45-20, 45-7) until the last block
(45-19), i = i + 1, i-th block
Shift the lock (45-22, 45-5) so that i is the last block.
The process is repeated until it becomes a click.   As a result, the j-th block for the i-th block
As the similarity VijIs obtained. Note that Vij = Vji, that is, the i-th block
J-th block multiple tone progression similarity to
Double progression similarity of the ith block to the block
Are equal. Vii = 100.   Calculation result of the degree of coincidence of double tone progression between blocks in Fig. 39 ｛Vi
j｝ is used in the generation of the hierarchical data in Fig. 40.
You.   In FIG. 40, c is a count for calculating the hierarchical structure.
Hj contains the hierarchical structure identifier of the j-th block.
Is set. Hj is an integer value of 0, 1, 2,...
Take. This is a, a'b, b 'in ordinary expressions.
(See HIEi in FIG. 6). Fig. 40 Flow
Then, for a certain reference block, 100%
Matching blocks have the same hierarchical structure identifier as the reference block.
With a hierarchical structure identifier of the same value (even value) (46-10, 46-
11), blocks that match in the range of 70-100%
A block with a compound sound signal that modifies the progression of rock sound
And the value obtained by adding 1 to the hierarchical structure identifier of the reference block.
A hierarchical structure identifier is attached (46-12, 46-13). Ma
In addition, blocks with similarity of less than 70% are
Treated as a block with a hierarchical structure independent of locks
You. The first block of the song as the first reference block
Selected (46-2), this reference block and 100%
Or for blocks that match 70-100%
The evaluation values of Hj = 0 and Hj = 1 are attached, indicating that the evaluation is completed.
Therefore, the flag flj of these blocks is set to “1”.
You. In this first evaluation loop (46-2 to 46-15)
The youngest block among the blocks of the song that were not confirmed
Becomes the reference block in the next evaluation loop (46-
3, 46-4, 46-6), the hierarchical structure of this reference block
Besshi is 2. Hereinafter, the same processing is repeated. Conclusion
As a result, the hierarchical structure identifier Hj is assigned to every block of the song.
Will be attached.   The processing in FIG. 41 is based on the block unit hierarchy obtained in FIG.
This is the process of converting the structure into hierarchical data in bar units.
You. That is, the HIEa in FIG.
The hierarchical structure identifier is set.   Next, referring to FIG. 42 and FIG.
Process to extract the pitch and set the scale key of each section.
Will be explained.   The concept behind the processing is as follows. (A) The tonality is the same rather than changing frequently in the progress of the song
It tends to maintain tonality. (B) Sounds other than the tones of the previous tones
When included, it should be transposed. (C) In the case of transposition, it changes to the close relatives such as F genus and genus.
Is desirable. (D) For any key that has a total of 12 different tones
However, if it contains sounds other than scale notes,
Use a key that maximizes the degree of coincidence with the scale.   In FIG. 42, the SC of 48-1 has an initially set sound.
Scale data (standard key, for example, scale data with C as the tonic)
Is set. This SC is used to determine the key of the current section.
In this case, it is used as the scale data of the immediately preceding section. 48−
Initializes the double tone counter i at 2 and scale data at 48-3
Initializes the counter c indicating the number of times that
With the key of the standard, it is transposed to the subordinate side (completely 5 degrees up
Substitute SC into a as the initial value for
Initialization for transposing to the lower subordinate side for the key of
To substitute SC for b.   48-5, whether the ith compound is a subset of scale a
Check if. For example, double tone data CC and scale data
If the result of logical AND with a matches the double tone data CC
For example, the double tone CC is a subset of the scale a. This condition holds
Substituting a into SC and the scale of the current section
A is used as the data SCALEi (48-13). Similarly,
In 48-6, it is determined whether the ith compound is a subset of scale b.
Check and if satisfied, substitute b for SC and SCALEi in 48-7.
Enter. If none of the conditions are satisfied, a
Turn b upward (48-7) and turn b completely down 5 degrees (48
-8), the rotation counter c is incremented (48-
9) Repeat the key determination. There are a total of 12 tones
The loop from 48-5 to 48-10 may be repeated six times.
That is, c> 6 is satisfied at 48-10 because of the duplication of the current section.
When the sound includes a sound that is not included in any scale
You.   In this case, the processing of 48-11, that is, as shown in FIG.
The processing is executed. In this process, the scale of the 12 keys
In the key that maximizes the degree of coincidence between the scale and
Seeking a scale. That is, for each key j = 1 to 12
Where j is the key scale and the constituent sounds of the compound
Is calculated (49-3), and the maximum degree of coincidence is set to max.
And find the key j that gives the highest degree of coincidence as P
(49-6). Finally, the scale data SC of the immediately preceding section is P times
The converted scale data is used as the scale data SCALEi of the current section.
This scale data is used to determine the key of the next section.
(49-9).   When the key of the current section is determined (48-13, 48-14, 48
−11), the counter i of the double tone is incremented.
(48-12), end of all double tone progression data (48-15)
Until the key determination process is repeated.   For example, diatonic scale
(Doremifasoraside) is selected and the previous section
Is C, and a plurality of current sections are D and F.
In this case, (C, D, E, F, G, A, B) is displayed in SC.
Contains data to be forgotten. Therefore, in FIG.
Then, after the processing of 48-3 and 48-4, the compound tone is scaled at 48-5.
a (this is the same as SC)
At 48-13, as the scale of the current section,
A diatonic scale is adopted. From this example
In this way, all of the constituent sounds of the
When it is on the scale of, the key of the current section is the same as the previous section.
Will be maintained.   Instead of D and F, D and F #
When used, a is the key in 48-3, and after 48-5.
Due to the diatonic scale of C, the double tone is a part
It is not a set, and this a is turned up 5 degrees by 48-7, and G
After obtaining a diatonic scale of 48-5,
When the sound is performed, the multiple sounds are a subset of a. And
The diatonic scale of G is the scale of the current section. This
As you can see from the example of
Contains sounds that are not in the scale of the previous section,
Transposition from key of this standard to close relatives, considering key between
Is repeated (for example, C → G → D → A ...)
The first key, which is the scale that contains all the constituent sounds of the double tone, is the current section
It is adopted as a key.   As described later, the scale data SCALEi of each section
The pitch organization of the melody to be used in the section to be used is determined. Sand
The melody sound of each generation section is determined by the scale data SCALEi.
Is selected from among the scale notes. <Melody generation>   The device according to the present embodiment is used for externally composable data.
And internally analyzed and evaluated the basic data
Then, it moves to the melody generation work.   Melody shown in 4-8 of the composition general flow (Fig. 2)
FIG. 44 illustrates a schematic flow of the key generation. Fig. 44 Smell
HIEi is the floor for each polyphonic section extracted in the previous polyphonic progress evaluation.
This is the layer structure data, as shown in 53-2 to 53-4.
Hierarchical data HIEi is the first stage melody type (LL)
It is used for generation control. More on this control later
Will be described. Further, the gradation structure data is shown in 53-5 and 53-6.
It can also be used to control the range of the sound type (LL).   The first melody type is the basic melody sequence.
Forming a skeleton, the range of the first stage melody is the sound of the melody
Regulate the area basically. In this embodiment, the control is
It uses the hierarchical structure obtained from the rows, which is also a feature of the embodiment.
It is one of the signs. But of the first stage melody
Hierarchical information extracted from polyphonic progression of tone control factors
It is not necessary to limit only to
Weight each of them, and calculate the first stage
Controls the sound type of the lodi, and specifies the weight by the user
You may be able to. In short, on the user's composition
It can be transformed so that the intention is reflected in the generation of the sound pattern.
Wear.   The melody of the first stage melody (LL) is the compound sound progress data CCi
Is used to express the pitch, that is, melody
(53-7).
Then, for the first stage melody, the second stage melody
Sound is added according to the production rules (53-
8). The product used to add this second stage melody sound
Action rules are the second melody included in the motif
It is the same as that used to classify the sound, and therefore
Therefore, reversibility is established for the analysis and synthesis of the melody.   Adding a second stage melody to the first stage melody
Thus, the pitch sequence of the melody is completed. Melody pitch
After the line is completed, the melody's length sequence (rhythm pattern)
Generation (53-9). Here, it consists of a predetermined number of notes.
Basic rhythm pattern (Essence generation 4-4 or E
Length sequence determined in the key extraction 4-6)
It depends on the pulse scale selected in setting 4-1.
Is changed into a pitch sequence with the same number of notes as the pitch sequence of the melody.
Is replaced.   First stage melody generation, second stage melody addition,
In the case of the flow shown in Fig. 44,
It is executed every time. Therefore, it is generated in a certain section in 53-10
The melody data is moved to the continuous area.   Figure 45 shows the generation, saving, and
44 is a detailed flowchart of the code (53-2 in FIG. 44,
53-3, 53-4). In this example, the hierarchical structure HIEi
The sound type LL is controlled as follows. First, wear
The section you are looking for is a section that is structurally different from the past
Whether or not the hierarchical data of passages
Is determined by comparing with the hierarchical structure data. Strange
Only for passages that are recognized as sections with
Generate type LL. This generation, the feature parameters
This is performed based on the data PC. Identify intervals with similar structure
No new sound pattern LL is generated for the music section. Instead
Of the sound patterns generated in the past,
Uses sound patterns generated in sections with the same structure as between
I do.   For example, assuming a song consisting of four passages,
If the structure of these four passages is a, b, c, a, respectively
I will try. The structure a of the first passage is a structure not seen in the past
Therefore, the sound pattern is generated according to the sound pattern feature parameters.
It is. Similarly, the second and third passages are new
Sound patterns are generated independently because
You. However, the final passage has the same structure a as the first passage of the song
is there. Therefore, as the sound pattern of the final passage,
Use the sound pattern generated for the verse as it is.   Generate a new sound pattern for a phrase with a new structure
That means another motif from this new structured passage
That occurs. Now in the first bar of the passage
The generated sound pattern is applied to the following measures of the passage.
When used repeatedly, a motif one bar long
Let's be aware. Generally, the length of a motif is one to several measures.
The length of the motif often changes in the middle of the song.
In the example of Fig. 45, taking this into account, the new structure
Motif length in the passage when is detected
One or two measures can be used. 2 bars
When the motif is selected, the first measure and 2
Independent measure is generated in bar no.
Measures of the first measure use the pattern of the first measure, and
Measure uses the note type of the second measure.   Reference to hierarchical data in the past section, past
Sound pattern (LL) data for repeating sound patterns in a section
A buffer is provided. Example of sound data buffer No. 46
Shown in the figure.   Explaining in accordance with FIG. 45, the passage at 54-1 (see FIG. 39)
Set the bar counter in the section indicated by each barno to "1"
I do. At 54-2, the hierarchical structure data HIEi of the current bar is
Measure Hierarchical Data HIE_iBy comparing with -1,
Check whether the passage has started. For example, | HIEi-HIE
_iThe time when −1 | ≧ 2 holds is the start of the passage. Passage
If it is determined to be the start, set the bar counter in the passage to “1”
It is reset (54-3). Then the sound data buffer
Search to see if the current phrase should generate a new note type
(54-4). Search sound type data buffer
Is performed as follows. First, the sound data buffer
Data at address 0 of the
Data indicating whether the address is
To the address of the data indicated from N to N
(Header information of each sound type) in sequence and read the top 8
The hierarchical structure data shown in the
Compare with Ta HIEi. The floor of the current measure in the sound data buffer
Hierarchical structure data similar to layer structure data HIEi (for example, HI
Same value as Ei or (HIE_iData with a value of -1)
When the current measure is
It is a bar. A header with similar hierarchical data is found
If this is the case, the data of the following sound pattern is
And load. For a new passage, start with a motif
Decide how many measures will be composed (54-5). This decision
Can be realized by random number generation, for example. 2 bar motif
The flag fl is set to "1" (54-7, 54-
8) Sound pattern for the next bar (the second bar of the passage)
Is newly generated. And generate the sound pattern
(Refer to Fig. 47), save in the sound pattern data buffer (54
-9). For details, HIEi x 0100 + measure counter value x 00
Create a header based on 10 + motif measures
The number N of tone sets at address 0 of the buffer
Increment and write header address to address N
From the address, the header, the LLNO
Length), LL₁, LL_Two…… LL_LLNO(Generated sound data)
Write. After that, other melody generation processing is performed.
(54-10), the bar counter is incremented (54
−11).   In 54-2, it was recognized that the passage was not changed
At this time, the flag fl is checked (54-13). fl = 1
Sometimes, the current measure is a phrase 2 that generates a two-bar motif.
Since this is the second measure, the sound pattern is generated again and the sound pattern data
And reset the flag fl to “0”.
(54-16). When fl = 0, the hierarchical structure data of the current bar
The header corresponding to the data HIEi is searched from the buffer, and
When the length information of the motif indicated by the header is 1 bar
Loads the following sound pattern data into its header and
When the bar length is 2 bars, the remainder of 2 in the bar counter is
Compare the measure numbers in the header, and if they match, add
Loads the subsequent note data as the note of the current measure.   First stage melody executed at 54-9 and 54-15 in FIG.
FIG. 47 shows the details of the generation of the sound pattern of the sound. The ckno of 56-1 is
Represents the number of constituent sounds of the compound sound. The number of polyphonic sounds is
Count the number that is "1" in the 16 bits of data CCi
Obtained by: In the example of Fig. 47, PC₁~ PC_FiveGenerate
This is a sound-type control parameter. r₁Is in the range of 1 to ckno
Means the number of the compound sound (56-
4). r_TwoIs PC_Three(Lowest sound of sound type) ~ PC_Two(The highest sound of the sound type)
Octave of LL
Indicates the node number (56-6). a = r₁+ R_Two× 0100
A candidate LL to be generated is calculated (56-7), and this candidate a is
When the condition of PC is satisfied, candidate a is adopted as LL
(56-8, 56-12, 56-14). However, depending on the value of PC
Is the preceding LL (for example, the first LL₁) Was decided
Followed by LL_TwoNo longer satisfies PC requirements
there is a possibility. For example, PC_Two= 501 (the highest note of the sound pattern is the fifth
1st double tone of octave), PC_Three= 401 (minimum of sound type)
The sound is the first double tone of the fourth octave), PC_Four= 3 (next
The maximum value of the difference between the LLs that match each other is three double tone constituent sounds), PC_Five= 3
(The minimum value of the difference between adjacent LLs is three sound components)
To LL₁As LL₁= 403 (The first LL is the fourth octave
Assuming that the first polyphonic sound was obtained, PC_Four, PC_FiveOn the condition
To match, LL_Two= 503 or 303 (with 3 sounds
This is a PC_Two, PC_ThreeDoes not meet the conditions. others
Prepare a loop counter LOOPC as a preventive measure
Strong when the counter LOOPC exceeds a certain value (for example, 100)
Candidate a is regularly adopted as LL (56-9, 56-1
0, 56-11).   FIG. 48 shows the details of the check 56-8 in FIG. 47. LLi
In order for candidate a to satisfy the condition of PC, (B) a ≦ PC_Two(Below the highest note) (B) a ≦ PC_Three(Must be higher than the highest note) (C) | a-LL_i−1 | ≦ PC_Four(The difference from the previous LL is the maximum value PC_Four
Below) (D) | a-LL_i−1 | ≧ PC_Five(The difference from the previous LL is the minimum value PC_Five
Above) Must be established. In the flow of Fig. 48
If the condition is not satisfied, set the flag OK to “0”
(Olda in the figure represents the immediately preceding LL; see 56-13).   FIG. 49 shows details of 53-7 in FIG. The purpose is (O
1st step menu indicated by
The format of Lody's sound type LL is converted to
Melody pitch data indicated by tab number + pitch name number)
And store it in MEDi. 58−
The processing of 5, 58-6 is performed by the compound sound number of LLi (LLi∧00f
f) is greater than the number of notes (CKNO) of the compound of the current section
At the time, the compound sound number of LLi
Is the process of changing to the highest compound sound number
You. In the figure, c is a counter of the constituent sounds of the double tone, LLi∧ff00 is L
Li is an octave number, and j is a note name counter.   FIGS. 50 and 51 show the second stage menu at 53-8 in FIG. 44.
It is the details of addition of a loddy sound. The purpose of this process is the first stage
Add the desired second stage melody sound to the floor melody
It is to complete the pitch sequence of D. Second stage melody
Features of the above-mentioned second stage melody for adding sound
｛RSi｝, scale data with tonality obtained from the evaluation of polyphony 進行 S
CALEi｝ expresses the knowledge to classify the second stage melody sound
Production rules are used. Second added
The step melody sound must satisfy the following conditions. (B) Sound within the specified range (B) The scale sound indicated by the scale data obtained in the evaluation of polyphonic progression
That (C) The sound is not a compound sound (D) Conclusions obtained by the production rules and the planned
Match with the two-step melody sound identifier RSi   In FIG. 50, the loops outside of 59-4 to 59-18 are total
Repeat for the number of drawn second stage melody sound identifiers RSi
The loop from 59-5 to 59-16 is the first stage menu.
Repeat for the number of Lody notes. In 59-8 to 59-14,
From the lower limit lo to the upper limit up as candidates for the second stage melody sound
Each pitch data k within the range of the sound is inspected sequentially (No.
See Figure 52). The pitch data k is a scale sound,
At some point (59-8, 59-9), the function F is calculated (59
−10), execute forward-looking inference based on production rules.
(59-11), the second stage melody sound whose conclusion is planned
Check if it matches the identifier RSi (59-1
1). When they match, the pitch data k is in the second stage described above.
Satisfies all the conditions as a melody sound. But
To count the number of added second stage melody sounds.
Increments nctct and finds the second stage
The pitch data k of the floor melody sound is put into VMnctct,
Set the additional position j of the floor melody sound to POSTnctct, and
Consecutive flags fl_jTo “1” (59-19 to 59-2
2). In this example, at most one first melody
A two-step melody sound is added, and fl_j=
0 is the first stage melody sound MEDj and MED_{j + 1}The second stage during
Indicates that the melody sound has not been added yet.   Conclusion at 59-12 = When the condition of RSi is not satisfied,
The pitch data k of interest is the second stage melody sound
Is not satisfied, the pitch data k is incremented.
(59-13) and repeat the process. K at 59-14
> When UP is established, the first stage melody sounds MEDj and MED
_{j + 1}That the second stage melody sound was not added during
Means Therefore, j is incremented (59−
15) Second stage melody sound between next first stage melody sounds
Proceed to check whether or not can be added.   Details of setting candidate sounds in 59-6 are shown in Fig. 53.
You. In this example, the first and second stage melody sounds MEDi, MED
_{i + 1}5 semitones above the higher sound and 5 and a half lower than the lower sound
The pitch range to search down to the pitch is set (62-5 to
62-7). However, when i = 0, that is, paying attention
2nd stage before the first 1st stage melody sound of the section
When adding a melody sound, the first stage
The pitch range from 5 semitones to 5 semitones below the melody
(62-1, 2), when i = Vmedno, that is, paying attention
After the last stage 1 melody sound of the section
When trying to add a rody sound, the final first stage
The pitch range from 5 semitones up to 5 semitones below the melody tone
(62-3, 4).   Details of the calculation of F in 59-10 are shown in FIG. This example
Then, the 2nd stage melody sound is between the 1st stage melody before and after
Because only one is added to the
(F in the figure₁~ F_Three) Is set to the specified value.
You.   Details of the forward inference of 59-11 are shown in Figure 38 above.
You.   When the processing in Fig. 50 is completed, nctct is added.
The total number of the second stage melody sounds is stored, and i in the array {Vi} is stored.
First, the i-th added in the processing of FIG. 59
The pitch data of the two-step melody sound is stored, and the array ｛POST
The ith of｝ is added to the ith in the processing of FIG.
The position information of the obtained second stage melody sound is stored.   These data are stored in the melody
Is converted to the pitch sequence {VMEDi}. Array ｛VME
Di｝ is initially set to arpeggio {MEDi}. 60
-2 to 60-9 are arrays {POSTi}, {V
This is the process of sorting Mi｝. 60-10 to 60-19 smell
To the position indicated by the position data POSTi
The pitch data VMi is inserted.   In this example, the second stage melody sound is generated during the first stage melody sound.
Although only one rody tone can be added,
The processing may be changed so that a step melody sound can be added.
No.   At this point, the pitch sequence of the melody is completed. Remaining processing
Is the generation of a melody duration sequence.   FIG. 55 shows the flow of generating a melody pitch sequence (53
-9 details). First, obtained by essence generation or extraction
Increase the number of notes in the reference rhythm pattern in the section of interest
Number of generated notes Vmedno (number of melody pitch train data)
, And the difference a between them is calculated (66-1). Generate
If the number of notes is smaller than the number of notes in the reference rhythm pattern
(When a> 0), the maximum of the note by the pulse scale
The appropriate coupling is repeatedly executed by the number of differences (66-2 to 66-
6). The number of generated notes is better than the number of notes in the reference rhythm pattern.
When the number of notes is large (when a> 0), the optimal division of the
It is repeated for the number of minutes (66-7 to 66-11). In this example
16-bit data is used as the rhythm pattern data format.
Assign each bit position to each timing using
Indicates that sound is generated at the bit position with the value of “1”.
Finally, convert to MER data format (66-1
2).   Details of the combining process are shown in FIG. In the figure, PSCALEj is used
Represents the j-th component scale of the pulse scale to be used
RR is a rhythm pattern to be processed. RR bits
Is the point with the smallest pulse scale weight in “1”
The notes are combined by setting the note to “0”. For example,
Rhythm pattern , The normal no pulse scale (see Fig. 7)
)), The result of combining one note is Becomes   This is obtained as follows.   First, RR was initially 0001 0001 0101 0001 It is. On the other hand, the normal pulse scale is 1213 1214 1213 1215 It is. Normal pulse schedule in bit “1” of RR
The point with the lowest weight is the seventh point from the right end.
Position. The bit at this position becomes “0”. Accordingly
And the resulting RR is 0001 0001 0001 0001 And this is Is represented.   Details of the dividing process are shown in FIG. The split is a bit of the RR
Is the point where the pulse scale weight is the largest among “0”
Is set to “1”. For example, rhythm putter
N On the other hand, if you divide the note with the normal pulse scale
And the result is Becomes   Details of conversion to MER data format 66-12 are shown in Figure 58.
You. In the figure, c₁Is a note counter, c_TwoIs the sound of each note
This is a counter that measures the length. In this example, MERo has RR
Since the length is entered until the first “1” appears,
It can also process notes that straddle boundary lines (bar lines) (Shinko
Measures).   Details of the data movement 53-10 are shown in FIG. First, MERo
The last sound that has generated (the blank part at the beginning of the current generation section)
Add to note length data MELRmeldno. Here meldno is already
Represents the number of notes being generated. The pitch generated this time
Column VMED₁Move ~ VMEDvmedno to MELD
Long train MER₁Move ~ MERvmedno to MELR (70-2-70-
6). Update meldno and finish (70-7). <Summary of composer>   As is clear from the above description, the automatic composer
This device has various features, some of which are described below.
Shown in (A) First stage melody sound that forms the melody skeleton
Progression of multiple sounds (primary constituent sounds) to define the set of highs
Information is entered. (B) The melody pitch sequence is the first stage melody generation
Then, in a two-stage process of adding a melody in the second stage
Generated. (C) Using music knowledge in melody analysis and synthesis
A production system that performs inference
You. (D) Classify the second stage melody sound included in the melody
Processing and adding the second stage melody to the first stage melody
Production that expresses the same musical knowledge
Because it is based on rule data, it can be used for both processes
There is inversion. (E) Analyze the progression of the double tone given as the composition material
By extracting the hierarchical structure and tonality structure of the song,
Is planned. (F) The extracted tonal structure is the key of the scale used in each section.
Stipulate. As a result, the tune controlled in tonality
Is guaranteed. (G) The extracted hierarchical structure is the sound of the first stage melody.
It is used for composition control. This allows for a variety of generated songs
Can incorporate gender and consistency.   The above-described embodiment is merely an example, and various modifications may be made.
Shapes, modifications and improvements are possible. For example, the embodiment described above
Then, the feature PC of the sound pattern extracted from the motif is used to create the sound pattern LL.
Used as control data in the
The sign PC may be changed as the music progresses. This is an example
For example, calculation of a function using the progress position of a song or the hierarchical structure as a variable
It can be realized by means.   Similarly, regarding the feature of the second stage melody {RSi}
Can also be changed as the song progresses. example
For example, the characteristics of the second stage melody extracted from the motif
One second-stage melody sound identifier is transferred to another second-stage melody.
Replace with rody tone identifier. This is the second stage melody
Randomly select one from a set of sound identifiers
This can be achieved.   Further, regarding the rhythm, in the above embodiment, the pulse scale is used.
Control the rhythm by combining and dividing notes
But the dominant minirhythm putter in the motif
And extract it into the melody sequence
It may be incorporated. [The invention's effect]   As described above, the automatic working machine of the present invention provides a basic
Data in the motif and the generated melody line.
Used as the sound of the first stage melody
Input means for inputting the progress information of the type of pitch (primary constituent sound)
From the motif, and the motif and
The length of the motif, which is a predetermined part of the progression of the primary constituent sound
Is compared with the part corresponding to
First characteristic indicating how the next constituent sounds are distributed
Other than the parameter and the primary constituent sound in the motif
A second characteristic parameter showing how the sounds are distributed
The first stage melody generating means generates the first
Feature parameters are changed by random numbers, and this and each music
Consists of only primary constituent sounds based on the primary constituent sounds of the section
The first stage melody is generated, and the melody generating means
The melody of the first stage based on the second feature parameter
Sound other than the primary constituent sound to the melody
(Second stage melody). did
Therefore, the final melody is created by the first stage melody.
The melody of the second stage is equipped with this skeleton.
Works to decorate, connect and assist.   This automatic composer uses the given motif as the primary constituent
The melody is generated by interpreting from the angle of the line, and the primary
Depending on what progression is given as the component sound progression
You can plan different songs. Accordingly
A wide range of compositions, from harmonious to non-harmonic
It is possible. Also, enter the chord progression from the top
Users need knowledge of harmony as there is no need to
And not.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is an overall configuration diagram of an automatic composer according to an embodiment of the present invention, FIG. 2 is a flowchart showing an overall operation in a composer mode, and FIG. Figures showing a list of the main variables used, FIG. 4, FIG. 5, FIG.
7 and 8 are diagrams showing the format of data used, FIG. 9 is a flowchart for initialization, and FIG. 10 is an example of compound sound progression data stored in a compound sound (primary sound) progression memory. FIG. 11, FIG. 11 is a flow chart of reading the multi-tone progression data, FIG. 12 is a diagram exemplifying pulse scale data stored in the pulse scale memory, FIG. 13 is a flow chart of reading of pulse scale data, FIG. FIG. 15 is a diagram exemplifying production rule data stored in a rule data memory, FIG. 15 is a flowchart for reading production rule data, FIG. 16 is a diagram exemplifying melody data (motif data) stored in a motif memory, FIG. The figure shows a flow chart for reading the melody data, FIG. 18 shows a flow chart for generating the essence, and FIG. Flowchart for setting the characteristics of the stage melody sound type, FIG. 20 is a flowchart for setting the characteristics of the second stage melody, the flowchart FIG. 21 to evaluate the motif rhythm each section, FIG. 22 Ps, P
e, Pss, Pee calculation flowchart, FIG. 23 shows Ps, Pss
24 is a flowchart for calculating Pe and Pee, FIG. 25 is a flowchart for extracting a first-stage melody tone of a motif, and FIG. 26 is a flowchart for extracting a feature of a first-stage melody tone. Figure 27 shows the second
FIG. 28 is a flowchart for extracting characteristics of a step melody, FIG. 28 is a flowchart for classifying a melody sound into a first-step melody sound and a second-step melody sound based on compound sound data, and FIG.
Figure is a flowchart for calculating the function F representing the status of the analyte melody, the flow chart of FIG. 30 calculates functions F _1, the flow chart of FIG. 31 calculates the function F _2, FIG. 32 calculates the function F ₃ of the flowchart, FIG. 33 is a flowchart of the calculation of the function F _4, FIG. 34 is a flowchart of the calculation of the function F _5, FIG. 35 is a flowchart of the calculation of the function F _6, 36
FIG. 37 is a flowchart for calculating the functions F ₇ and F ₈ , FIG. 37 is a flowchart for temporarily storing the calculated function F, FIG. 38 is a flowchart for inferring the type of the second-stage melody sound, and FIG.
FIG. 39 is a flowchart for calculating the degree of coincidence of the polyphony progression between blocks, FIG. 40 is a flowchart for generating hierarchical structure data from the calculated degree of coincidence, and FIG. 41 is a diagram showing the hierarchical structure data of the block for each polyphonic section 42 and 43 are flowcharts for determining the key and scale of each section from the progression of multiple sounds, FIG. 44 is a flowchart of melody generation, and FIG. 45 is a first-stage melody tone generation. Flowchart showing save and load,
FIG. 46 is a diagram exemplifying a sound pattern data buffer, FIG. 47 is a flow chart of sound pattern generation, FIG. 48 is a flowchart of check in sound pattern generation, and FIG. 49 is a flow chart of converting a sound pattern to a melody data format. 50 and 51 are flowcharts for adding a second-stage melody sound to a first-stage melody, FIG. 52 is a diagram showing an order of a second-stage melody sound adding process, and FIG. 53 is a second-stage melody sound. 54 is a flowchart for setting a range of pitches as candidates for the function F. FIG.
55 is a flowchart for generating melody note length data, FIG. 56 is a flowchart for optimal combination of notes, FIG. 57 is a flowchart for optimal note division, and FIG. 58 is a flowchart for the generated rhythm pattern. FIG. 59 is a flow chart for converting the generated melody data to a continuous area. 1 ... CPU, 2 ... input device, 3 ... motif memory,
4: Multitone progression memory, CCi: Multitone data, LLi: First-stage melody type, PSi: Features of the second-stage melody, PCi: Features of the first-stage melody sound type.

Claims (1)

(57) [Claims] Motif input means for inputting a motif; primary constituent sound progression providing means for providing a primary constituent sound consisting of one or more note names for each music section as a primary constituent sound progression over a plurality of continuous music sections; A first characteristic parameter indicating how the primary constituent sounds are distributed in the motif by comparing a predetermined part of the primary constituent sound progression with a part corresponding to the time length of the motif; And a second characteristic parameter for generating a second characteristic parameter indicating how the sounds other than the primary constituent sounds are distributed in the motif; and changing the first characteristic parameter by a random number. A first-stage melody generating means for generating a first-stage melody composed of only the primary constituent sounds based on the primary melody and the primary constituent sounds of each music section; A melody generating means for adding a sound other than the primary constituent sound to the melody of the first stage in accordance with the characteristic parameter of and outputting the melody as a melody.