JPH01167882A - Automatic musical composition machine - Google Patents

Abstract

PURPOSE:To widely execute a musical composition extending from a harmonious melody to an unharmonious melody by forming a framework of a final melody by a melody of a first stage which is generated, decorating this framework by a melody of a second stage, linking and assisting it. CONSTITUTION:From an input device 2, a motif, a compound sound advance, a kind of a pulse scale to be used, and a kind of the musical scale are inputted. Subsequently, in a memory 3, the inputted motif is stored, and in a memory 4, compound sound advance information is stored and by a CPU 1, the first stage memory is extracted, etc. In such a state, a user can select the musical scale to be used from a memory 5, and selects the second stage melody from a production rule data memory 6. Next, at the time of starting a musical composition, the user selects a pulse scale from a memory 7 in order to generate a rhythm of the melody, and completed melody data is stored in a memory 8. In such a state, in monitor 11, a result of musical composition is displayed and outputted through a CRT 12, a staff notation printer 13, a musical sound forming circuit 14 and a sound system 15.

Description

[Detailed description of the invention] [Technical field of invention] This invention relates to an automatic music composer.

[Background] The sounds that make up the melody line of ordinary music can be divided into harmonic sounds and non-harmonic sounds. Focusing on this point, the applicant filed the patent application No. 86571 (April 1986).
(Application filed on August 8th) extracts the characteristic parameters of dispersive chords and the characteristic parameters of non-harmonic sounds included in the motif from the chord progression of a given motif, and according to the extraction results, constructs the chord in each melody generation section. We have filed an application for an automatic composing machine that uses sounds to generate a sequence of dispersive chords, and then adds non-harmonious tones to complete the sequence of melody notes.

Therefore, while this automatic composer can be used to compose harmonically pleasing music, the chords for each melody generation section must be selected from a set of chords built into the automatic composer. . This places harmonic constraints on the generated melody.

Further, in order for the user to input a chord progression, knowledge of harmony is required as prior knowledge.

[Purpose of the Invention] This invention was made in view of the above-mentioned background, and its purpose is to enable a wide range of compositions from harmonic to non-harmonic songs, and to create a system that is exceptional in use. To provide an automatic composition machine that does not require musical knowledge.

[Structure and operation of the invention] In order to achieve the above object, the present invention uses a motif (one to several bars of melody) input by the user and the primary constituent sounds (the sounds included in the first stage melody). a feature parameter extraction means for extracting the feature parameters of the first stage melody and the feature parameters of the second stage melody based on the progression of the melody generation section; a first stage melody generating means for generating a first stage melody according to the characteristic parameters of the first stage melody; It is characterized by having a second stage melody adding means for adding a melody.

The above-mentioned primary constituent sounds are composed of one or more sounds.

This automatic music composer generates melodies in two stages.
The melody of the first stage constitutes the framework of the melody that is finally generated, and the melody of the second stage serves to decorate or connect this framework.

This double-structured melody line is the primary constituent sound,
Depending on what kind of sound you choose, the songs change in a variety of ways, sometimes creating traditional songs, and other times creating new, experimental songs.

[Examples] Examples of the present invention will be described below. In composer mode, this automatic composition device adopts an approach of composing sounds by dividing them into first-stage melody sounds and second-stage melody sounds. Data that forms the basis of composition include progression information of primary constituent tones (hereinafter referred to as complex tones) that determine the pitch organization of the first stage melody tones, motifs (melody input by the user), and the rhythm or rhythm of the melody to be generated. The pulse scale used to control the note length sequence and the types of basic scales are given. Each sound included in the motif is distinguished into a first stage melody sound and a second stage melody sound by using the compound sound information assigned to each section. The part obtained by removing the second-stage melody tones from the motif is the tone sequence of the first-stage melody tones of the motif. From this string of notes, the positional features of the first stage melody are extracted. Furthermore, for a motif that has been identified as a first-stage melody sound and a second-stage melody sound, musical knowledge for classifying the second-stage melody sound (this is stored in the production rule data memory described later) is used. ), it is possible to extract the type (characteristic) of each second stage melody sound included in the motif.

That is, information (characteristics of the second stage melody) indicating how the second stage melody sounds are distributed within the motif can be obtained. Furthermore, the hierarchical structure and tonality structure of the song are extracted from the multitone progression.

The generation of the melody consists of the first stage melody generation process, the second stage melody generation process, and the second stage melody generation process.
It consists of a stage melody addition process and a tone length sequence generation process.

In the step of generating the first stage melody, the generation of the position type of the first stage melody is controlled by the above-mentioned hierarchical structure extracted from the multitone progression information.When the generation of the position type of the new first stage melody is instructed by the zero layer structure. , the first stage melody position type is generated from the first stage melody position type feature (the first stage melody position type feature extracted from the motif or a modification thereof), and the first stage melody position type is By applying the corresponding polytone information, a first stage melody expressed by a series of pitches is generated. A second stage melody is added to the first stage melody thus generated. The musical knowledge described above is again used to add the second stage melody. In the inference of second-stage melody sound addition, the conditions are that the sounds that can be added satisfy the characteristics of the second-stage melody and are scale sounds. The scale tones here are tones on a scale obtained by inverting the tonic (key, tonality) of the basic scale according to the tonality structure extracted from the multiple-tone progression. In the classification of second stage melody sounds and the addition of second stage melody sounds,
Because inferences are made using common musical knowledge, ``reversibility'' is provided between the analysis and synthesis of melody by the device. Complete reversibility means that by analyzing a certain melody,
If a certain analysis result is obtained, and conversely a melody is generated based on the analysis result, the generated melody will match the original melody. By adding the second stage melody to the first stage melody, the pitch sequence of the melody is completed. On the other hand, the tone length sequence of the melody can be obtained as follows. That is, the basic rhythm (sequence of note lengths) is optimally combined or divided using a pulse scale so that the target number of notes (for example, the total number of harmonic and non-harmonic sounds) is achieved. How many notes are divided or combined depends on the weight of each pulse point of the selected pulse scale. Therefore, consistent rhythm control is possible.

Overall Configuration> FIG. 1 shows an overall configuration diagram of the automatic composing machine according to this embodiment. The CPUI is a control device for realizing the automatic composing function in this apparatus. From the input device 2, the motif (melody), multitone progression, type of pulse scale to be used,
The type of scale to be used, etc. is input. The motif memory 3 stores the input motif. The multiple note progression memory 4 is a memory that stores multiple note progression information.
The information is used by the CPUI when analyzing a polyphonic progression, when extracting or generating a first stage melody, etc. The scale data memory 5 is a memory that stores scale data representing various types of scales.
The user can select a specific scale from the set of scales in the memory 5 as a scale to be used in a song. The production rule data memory 6 stores musical knowledge for classifying the second stage melody sounds. This music knowledge is used when classifying the second stage melody sounds included in the motif and when adding the second stage melody to the generated first stage melody. Pulse scale memory 7 stores various pulse scales (set of pulse scales)
When starting a composition, the user can select a desired pulse scale from this set of pulse scales, taking into account the rhythmic signature of the song. The pulse scale is used to generate the rhythm (sequence of note lengths) of a melody.

The completed melody data is stored in the melody data memory 8. The external storage device is used as a copy of the melody data stored in the melody data memory 8, other music knowledge, and a resource for another composition program. work memory l
O stores various data used by the CPU during operation, such as tonality structure, hierarchical structure, and various variables. The monitor 11 is composed of a CRT 12, a staff notation printer 13, a tone forming circuit 14, and a sound system 15, and can display and output composition results and analysis results through these devices.

General Composition Flow> Figure 2 shows the overall flow (general composition flow) of this device in composer mode.

In the initial setting 4-1, the user inputs basic information for composing music to the automatic composing machine. The information that the user informs the automatic music composer is (1) BEAT, (2)
) Type of pulse scale (3) Scale (4) Fully automatic or motif input is shown. BEAT is the length of one bar expressed in the number of notes (the shortest note) of the basic unit length,
It is what defines the time signature of the song. For example, 4
For songs with time signatures, if the basic unit length of notes is a 16th note, then if BEAT=16 is set, the length of one bar will be 4 beats. The pulse scale selected in step 4-1 is information that primarily controls the rhythm of the music composed by the automatic music composer. A pulse scale is a scale in which each pulse point, which has an interval of the basic unit length, is given an easterly edge that indicates the ease with which notes can be combined or divided.
12), the tone length sequence of the melody is controlled by using this pulse scale.

Therefore, selecting the type of pulse scale corresponds to selecting the rhythmic characteristics of the piece of music to be composed by the automatic composer. The type of scale selected in step 4-1 (for example, diatonic scale) is a scale used by the automatic composer in composing music. Further initial settings 4-
1, the user decides whether to compose the music fully automatically or based on a motif. When fully automatic, the data necessary for composing (1) multitone progression, (2) production rules, and (3) pulse scale are loaded into the work memory lO (4-3), and (1) the basic rhythm. (note length pattern), (2) characteristics of the first stage melody pattern (PCi: see Figure 5), and (3) characteristics of the second stage melody (R5i: see No. 55!J). Generated according to the instructions (4-4), −
On the other hand, in the composition mode that uses motifs, the motif (input melody) is also read in addition to the above data as data reading 4-5, and the essence of (1) rhythm (2) first stage melody pattern (3) first stage melody pattern is read. Characteristics of the step melody pattern, (4) The characteristics of the m2 step melody are:
(4-5) Extracted from this motif, in both fully automatic and motif usage modes, evaluation of polyphonic progression 4-
7 is performed, where (1) hierarchical structure (2) tonal structure is extracted from the polytone progression information. This hierarchical structure expresses the coherence and diversity of the song inherent in the multitone progression. Further, the tonality structure defines the key (fundamental tone) of the scale used in each section. By processing steps 4-7,
For example, the “analytical work” for composing has been completed.
The characteristics of the first stage melody pattern are information necessary to generate a pattern of the first stage melody, and the characteristics of the second stage melody characterize the second stage melody added to the first stage melody. In addition, the tonality structure has a new 0-layer structure that constrains the candidate notes of the melody for each section.
It can be used to decide whether to generate a stage melody type. The pulse scale is also used to generate rhythm. In the melody generation 4-8, (1) generation of a first stage melody, (2) addition of a second stage melody, and (3) generation of a rhythm are executed.

Variable list, data format> Figure 3 shows a list of the main variables used in the flow described later, and Figures 4 to 8 show the data formats.The data formats are merely examples, and other arbitrary data formats are shown. Any suitable data format can be used.

The composer mode of this headpiece will be explained in detail below.

Initial settings> Initial settings in the composition general flow (Figure 2) 4-1
The details are illustrated in FIG. Select B with this initial setting.
The meaning of EAT, the pulse scale type (PULS), the scale type (ISCALE), and the meaning of fully automatic or motif input have been explained with reference to FIG. 2, and will therefore be omitted here. The value of PULS is stored in pulse scale memory 7 (first
The value of I5CALE serves as a pointer to a specific type of pulse scale stored in the scale data memory 5.
This is a pointer to a specific type of scale data stored in the .

Data Reading> As shown in the composition general flow of FIG. 2, after the initial settings, data reading is executed in 4-3 or 4-5. In the case of full automatic mode, motifs are not used as basic data for composition, so motif data is not loaded.The loading of each data will be explained below.

FIG. 10 shows an example of the multiple note progression data stored in the multiple note progression memory 4 (FIG. 1), and FIG. 11 is a flowchart for loading the multiple note progression data stored in the multiple note progression memory 4. In the memory map shown in Figure 10, the note data (CCi) constituting a compound note is placed at an even address, and the length of the note is placed at the next address (odd address). CCi with a value of 507
represents the four tones G, B, D, and F, and CRi having a value of 1O represents that the length of the compound tone is 16 times the basic time length (for example, a 16th note).

In Figure ti, the register CCi contains the i of the song to be composed.
The compound note data that appears the th time is entered, and its length is entered in CRi. CDN0 also contains the total number of double tones. Since other points are clear from the description in FIG. 11, their explanation will be omitted.

FIG. 12 shows an example of pulse scale data stored in the pulse scale memory 7 (FIG. 1), and FIG. 13 is a flowchart for loading the type of pulse scale selected in the initial setting from the pulse scale memory 7. In this example, the pulse scale [1(P
ULS) points to a specific address (e.g. 0) in the pulse scale memory 7, and at that address,
Contains the start address of the selected pulse scale data. This start address contains the subscales that make up the pulse scale (scales that only have weights of 0 and 1).
For example, the normal pulse scale has 5 subscales FFFF, 5555.1111.
．． 0101.0001 (1B representation), and the corresponding binary representation is shown in FIG. In the case of the normal pulse scale, the weight of the first pulse point (the rightmost position WI in Fig. 7) is the maximum, 5, which means that
This indicates that when the normal pulse scale is selected, a note is most likely to exist at the first position of each section (small m) of the generated rhythm.

FIG. 14 shows the production rule data memory 6 (first
An example of production rule data stored in Figure 1) is shown. FIG. 15 is a flowchart for reading data stored in this memory 6. The entire production rule is an expression of musical knowledge for classifying the second-stage melody sounds included in the melody, and each production rule data includes lower limit data Li as data that defines the prerequisite part of the rule, and a function It has function instruction data Xi indicating the type, upper limit data Ui, and data Yi and Ni as the conclusion part of the rule. The function is a numerical expression of the state of the melody to be analyzed, and an example thereof is shown in FIG. 29, which will be described later. The value Fxi of the function indicated by data Xi is greater than or equal to Li and less than or equal to Ui (Li≦Fxi
≦Ui) is the premise (life 8) of the production rule, and the conclusion when this premise holds is shown by data Yi, and the conclusion when this premise does not hold is shown by data Ni. There is. And data Yi or Ni
When has a positive value, that value indicates the next production rule number to be referred to in forward inference, and when it has a negative value, the type of second-stage melody sound is expressed by its absolute value. Ru. Forward inference always starts from one rule, and this rule is called a root, and forward inference ends when a conclusion Yi or Ni having a negative value is found.

In the case of the address assignment of the production rule data shown in FIG. 14, the data of each production rule is stored in five consecutive addresses starting from the address of the lower limit data Li. In detail, Li is assigned to addresses that are divisible by 5, Xi is assigned to addresses that have a remainder of 1 when divided by 5, and addresses that have a remainder of 2.
Data of Ui is stored at the address of , data of Yi is stored at the address of remainder 3, and data of Ni is stored at the address of remainder 4.

In FIG. 15, the total number of production rules is set in RULENO. Other points are clear from the above explanation and flow description.

FIG. 16 shows an example of motif data (melody data) stored in the motif memory 3, and FIG. 17 is a flowchart for reading the motif data that is the basis of a composition. In the case of Fig. 16, the pitch data M of the note is at the even address.
The tone length data MRf of that note is stored at the next odd address after Di. The number of notes of the motif is set in MDNO in FIG. 17.

This concludes the explanation of data loading.

Generation of Essence> In the fully automatic composition mode that does not use motifs, after reading the data, the rhythm, characteristics of the first stage melody pattern, and characteristics of the second stage melody are generated as the essence of the song (see 4-4 in Figure 2). 4).

FIG. 18 shows a flowchart of this essence generation. These essences can be specified by the user or generated completely automatically. For example, the reference rhythm pattern in 22-1 is set by automatic rhythm pattern generation means, for example, 4/4 time as the time signature and normal as the pulse scale. This can be done by automatically generating Tsuno'no as a reference rhythm pattern when the rhythm pattern is being played, or by inputting the user's favorite rhythm pattern. 22-2
The characteristic setting of the first stage melody pattern in 22-3 and the characteristic setting of the second stage melody in 22-3 are also performed automatically or by user input. FIG. 19 illustrates a flowchart for automatically setting the characteristics of the first stage melody pattern by random number generation, etc., and FIG. 20 illustrates a flowchart for setting the characteristics of the second stage melody pattern by input by the user.

In the feature setting of the first stage melody pattern (Fig. 19), PCl to PCs are the number of notes constituting the first stage melody within a section of a predetermined length (for example, a measure), the highest note, the lowest note, and the adjacent first stage. Each PC can generate the maximum value of the difference between melody tones and the minimum value of the difference between adjacent first stage melody tones (see Fig. 5) for each section.
It can be obtained by setting the upper and lower limits of the possible values and generating random numbers between them. Alternatively, you can prepare a database of PC series corresponding to the progression of the song, and select the desired PC.
It is also possible to select a series.

In the second-stage melody feature setting shown in FIG. 20, the monitor displays the keyword corresponding to the type a of each second-stage melody sound to prompt the user to input it (24-
2). The sequence of the second stage melody sound identifier a input by the user is entered in the array of R5i (24-3.24-6.24
-7), when calling the code EOI indicating the end of input,
The number of second stage melody sounds is entered into R3N0 and the flow exits (24-8).

Extracting the essence of the song> In the composition mode that uses motifs, after reading the data, extracting the essence of the song from the motif (rhythm, first stage melody pattern, characteristics of the first stage melody pattern, second stage melody pattern,
Characteristics of staged melody sounds) are extracted (Fig. 2, 4-6).
).

The rhythmic evaluation of the motif is shown in Figure 21, the first stage melody pattern extraction is shown in Figure 25, the first stage melody pattern (placed type) feature extraction is shown in Figure 26, and the second stage melody feature extraction is shown in Figure 27. An example is shown in the figure. In these flows, each essence is extracted for each section (measure).

In the rhythm evaluation in Figure 21, Ps at 25-1
contains positional information indicating the number of the first note of the target measure in the song, and PSS contains positional information indicating the number of notes in the song at the beginning of the target measure, and PSS indicates the length of time that the first note of the measure, indicated by Ps, protrudes into the previous measure (basic time expression) is entered, and Pe contains the position information of the last note of the target measure (the note one note before the first note of the next measure). rr shown in 25-2 is a 16-bit register that stores the rhythm pattern of the target measure.If the length of one measure is 16, the first bit position of register rr represents the first basic time of the measure. , similarly, the Nth bit position represents the Nth basic time from the beginning of the measure, 25-3
The processing from 25-9 to 25-9 uses the motif note length data MRi to determine the position of the note from the motif note Ps to the note re.
This is the process of calculating the value using , and writing it into the corresponding bit position of register rr. For example, as rr, ooo i
If the result of ooo i ooo t ooo t is obtained, this pattern rr will be the 1st and 2nd beats of the target measure.
This indicates that the sound is generated on the 3rd, 3rd, and 4th beats.

Details of calculation of PS, PSS, Pe, and Pee are shown in Figure 22~
It is shown in FIG. Fee represents the length by which the note following Pe, that is, the first note of the next measure, cuts into the target measure.

In the calculation flow of Ps and Pss in Fig. 23, bea
t is the length of one bar (basic time length expression), bar is the number of the target bar (represents the bar number specified by the user, the specified bar number is smaller than 1, but the number of bars in the song (mn
o) If it is larger than that, it is an input error. When the specified measure is 1, Ps=1. Make Pss Tomi O (27-4.
27-5), the reason why Ps = 1 is that in the case of the first measure, the first note of the measure is the first note of the song or the first note of all motif data, and Pss
=O because there is no preceding measure.

The al calculated in 27-2 is the length from the beginning of the song to the front bar m line of the target measure, and this length al is the length S obtained by accumulating the note length data MRi of the motif from the beginning. Compare with (27-7.27-8.27-10.27-12), 5
When =al holds true, the note next to the i-th pitch data accumulated last in S starts from the beginning of the target measure. Therefore, Ps= i +1, Pss=0 (2
7-9), On the other hand, when S>al holds true, the note with the note length data added last to S, that is, the i-th note, is the first note of the target bar. Therefore, Ps=
Let it be i. Also, let Pss=MRi-3+a1 (2
7-9).

The calculation flow for Pe and Fee shown in FIG. 24 performs processing very similar to that in FIG. 23. However, al contains the length from the beginning of the song to the bar line after the target bar, and other points will not be explained.

In the first stage melody tone pattern extraction flow shown in FIG. 25, the first stage melody pattern (LLi) is extracted from the motif of the target bar. To outline the process, P
For the target measure indicated by s and Pe, it is determined whether the motif data is a first stage melody note using the corresponding compound note information in the compound note progression information, and the note determined to be a first stage melody note is For this, data in the form of LL is obtained by checking the number of notes from the bottom of the compound notes.To be more specific, first, from the motif data, first note is the first note to be evaluated. (P s) and find the last note (Pe) (29-1) 8 Next, load the compound note data (29-2), then initialize the note counter i and the first stage melody note counter k (29-3.29
-4). The processing of 29-5 is the motif pitch data MDi
This is a process of converting the data into the same data format as the compound tone data CC. For example, the sound 'G' is converted to data mm with 'l' in bit position 7.

At step 29-6, it is checked whether the pitch data mm is a complex note. This can be determined by calculating the logical product (mmΔcc) of the pitch data mm and the compound tone data CC. In 29-7 to 29-13, bit "l" of pitch data mm of the motif is set in bit "l" of compound note data CC.
”, and as a result, the octave number of the note of the motif (MDiAffoo) is added to C to create the first stage melody pattern data LLk. At 29-15, the counter is added to the next note. LL is obtained by advancing i until the note number reaches Pe (29-16).The number of notes of the first stage melody of the target section is entered in LLNO of 29-17.

The characteristics of the placement type of the first stage melody in Fig. 26 are the extraction result (LL i ) of the placement type of the first stage melody in Fig. 25,
Derived from LLNO.

In the second-stage melody feature extraction shown in FIG. 27, a pattern of types of second-stage melody sounds distributed in the motif of the target section is obtained. Set the counter and note counter for the second stage melody sound (33-2.33-3), and if the note of interest is the second stage melody sound (33-4)
, a function F that represents the situation of the motif centered on that note.
is calculated, and forward inference is performed using production rules to find the type of second stage melody sound and substituted into R3j (33-5 to 33-8). By performing this second stage melody sound classification processing until reaching re, the array (R3
In j), the arrangement of types of second stage melody sounds of the motif of the target section is set. The total number of second stage melody sounds in the target section is entered in PSNO 33-11.

Details of the process for determining whether MDi shown at 33-4 is a second stage melody tone are shown in FIG.

This discrimination process involves the extraction of the placement type of the first stage melody (the second
In Figure 5), this process is similar to the process of determining whether the note of interest is a first-stage melody note (a constituent note of a compound note), and the note of interest is among the notes of a compound note in the target section. This can be determined by whether or not it includes the note name.

In the calculation of the function F in step 33-6, conditions or factors of the motif (melody) necessary for classifying the second stage melody sounds are calculated in forward inference. Specific examples are shown in FIGS. 29 to 37. In this example, the function F is: Fl: How many points of the note of interest (second-stage melody sound) the first-stage melody sound is located at ( Position of backward 1st stage melody note g!1) F2: How many positions before the note of interest is the 1st stage melody note located (position of forward 1st stage melody note) F3: Front 1st stage melody note Number of second stage melody sounds between the first stage melody sound and the rear first stage melody sound F4: Pitch difference between the front first stage melody sound and the rear first stage melody sound F5: The front first stage melody sound and the rear first stage melody sound Pitch distribution of the second stage melody sound with respect to the stage melody sound F6: Whether or not the pitch of the melody changes monotonically from the front first stage melody sound to the rear first stage melody sound Fl: Backward stage melody sound Pitch difference between the first stage melody note and the previous note (pitch progression to the backward first stage melody note) F8: Pitch difference between the front first stage melody note and the next note (forward The interval progression from the first stage melody note is calculated (Figure 29).In addition to this, information indicating whether the beat is weak or strong, and information that distinguishes the note length can be added to the set of function F. Since the flowchart for calculating the good 0 individual functions F is clear from the description itself, the explanation will be omitted.

The details of the forward inference in 33-7 are shown in FIG. 38. First, the rule number pointer P is set to "1" which indicates the root rule among the production rules.
(44-1). After that, the prerequisite part of the rule indicated by the rule number pointer P (Lp≦Fxp≦Up
) holds, and if it holds, the data Yp of the affirmative conclusion part of that rule is used as a pointer to the next rule, and if it does not hold, the data Np of the negative conclusion part of that rule is used as a pointer to the next rule. Use as a pointer to a rule. However, when the data yp and Np are negative values, the final conclusion has been reached, so the absolute value (-Yp, -
Np) is set in the conclusion register as the identifier of the type of second stage melody sound. According to the flow, 44-3 condition L p > F X p or 44-5 (7) condition Fxp
>Up holds, the premise of rule P Lp≦Fxp
≦Up does not hold, so the data Np of the negative conclusion part of that rule should be substituted into a44-4.44-6); otherwise, the premise holds true, so the data Np of the positive conclusion part of rule P is assigned to a. Data yp is entered (44-2), and this a is assigned to P (44-7). If P is a positive value, return to the next rule inspection as the primary rule number pointer, and if 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 second-stage melody sounds, in the calculation of the function F above, F+=1: The backward first-stage melody sound is the second stage melody sound of interest.
F2 = -1, which is one after the stage melody sound. The first stage melody sound is one stage before the focused second stage melody sound. F3 = 1: The sound between the previous and succeeding first stage melody sounds. The number is 1. Fa-8: The pitch difference between the preceding and following 1st stage melody sounds is 8. Fs = 2: The 2nd stage melody sound between the preceding and following 1st stage melody sounds is the 1st stage melody sound before and after. F6 = 1: Distributed in the middle of the pitches of the melody sounds. The melody's pitch changes monotonically from the front first stage melody sound to the rear first stage melody sound. F7 = 1: Backwards first stage melody sound. Assume that Fa = 7, which progresses to the melody sound with a pitch difference of l: From the front first stage melody sound, a pitch difference of 7 is obtained, and the data shown in Figure 14 is used as production rule data. Let's try to make an inference.

First, when P=1 (root), the premise 0≦F2
≦0 does not hold because F2=-1. therefore,
The negative conclusion part N1;3 of the root becomes a pointer to the next rule to be checked.

When p=3, the premise 0≦F,≦O of rule 3 does not hold because F1=1. Therefore, N
5=5 becomes the pointer to the next rule.

At p=5, the premise 0≦F4≦0 of Rule 5 does not hold because F4=8, so N5
=6 becomes P.

When P=6, the premise l≦F of Rule 6 also holds true because F6 =1. Therefore, data Y6=7 of the affirmative conclusion part of rule 6 becomes the pointer P of the next rule to be accessed.

When P=7, premise 3≦FB≦(1) of rule 7 holds true because F8=7. Here Y7 ==-2
(negative). Therefore, conclusion = 2 (classified second
(Identifier of step melody sound).

As can be seen from this example, for any type of second-stage melody sound, a finite number of propositions hold true (the fact that the premise part does not hold is equivalent to the proposition that the premise part is false) holds true. It is possible to identify it with the knowledge of music. To express this knowledge.

Function F has been calculated and production rules have been created. That is, the function F is information on melody check items used in the knowledge for classifying the second stage melody sounds, and the production rule data is a sequence of rules leading to the classification result of each second stage melody sound. are connected using pointers.

Compound Progression Evaluation> One of the features of the apparatus of this embodiment is that it makes maximum use of multitone progressions for composing music. In other words, in the multitone progression evaluation shown in 4-7 of the composition general flow in Figure 2, the 0-hierarchical structure, which uses the given multitone progression as a clue to determine the hierarchical structure and tonality structure of the piece, evaluates the consistency and diversity of the piece. It is used to control the generation of the first stage melody in melody generation, which will be described later. On the other hand, the tonality structure represents changes in tonality (key) as the song progresses, and is used to select the scale key to be used in each section in melody generation, which will be described later.

Details of the extraction of the hierarchical structure will be described below with reference to FIGS. 39 to 41.

The flow shown in FIG. 39 is a flow for calculating the degree of similarity of polyphonic progression between blocks, using blocks each having a length such as a passage as a unit.

First, the length of the song SUM is the length of each compound note in the compound note progression information C
Obtained by accumulating Ri (45-1), barno
(45-2) Convert the length of block C' −/ ri, which is expressed as (small y5 number), into the expression of basic tone length and make the block length again.45-2)
, Calculate the number m of blocks included in the song by dividing the acrobatics SUM by the block long sentence (45-3), Initialize the counter i for the reference number of the block to be compared to “0” (
45-4).

45-8 to 45-18, the i-th block and j
Similarity of diphthong progression with the th block (j≧i) Vij
is being calculated. As a function of similarity, is used. Here the sentence is block length and V
s represents the number of matches of compound tones obtained when comparing the compound tones of the i-th block and the compound tones of the j-th block for each standard tone length. This similarity function Vij takes a value from O to 100. , 100, the polytone progressions of the two blocks match completely (100%), and when O, they completely mismatch.

The calculation of the similarity between the i-th block and the j-th block starts from the position j=i (45-6), and each time a similarity is obtained, the similarity with the next block is calculated by j=j+1. (45-20, 45-7) until the final block (45-19), due to i=i+1,
Shift the i-th block 45-22.45-5)
, i repeats the process until it becomes the final block.

As a result, as the similarity vij of the j-th block to the i-th block, V l l V '+
2 ”” ”” ”” 9. Il+" -0000
,V l nV22self・・・・・・・・・・・・・・――■・−V
2n-'v n'n is obtained. Note that Vij=Vji, that is, the j-th block composite progression similarity for the i-th block,
The polyphonic progression similarity of the first block to the jth block is equal and vii=100.

Calculation result of compound tone progression matching degree between blocks in Fig. 39 (V
ij) is used in the hierarchical structure data generation shown in FIG.

In FIG. 40, C is a counter for calculation of the hierarchical structure, and the hierarchical structure identifier of the j-th block is set in Hj. Hj is a value of 0.1.2...
・Takes an integer value. This is a, a' in normal expression.
b, b'... (HIE in Figure 6)
i), in the flow of Fig. 40, a block whose 100% polyphonic progression matches a certain reference block is given a hierarchical structure identifier of the same value (even value) as the hierarchical structure identifier of the reference block (46- 10, 46-11), 70-1
A block that matches within the range of 00% is assigned a hierarchical structure identifier with a value of 1 added to the hierarchical structure identifier of the reference block, as a block with a multitone progression that modifies the multitone progression of the reference block (46-12.46 -13), and
Blocks having a similarity of less than 70% are treated as blocks having a hierarchical structure independent of the reference block. The first block of the song is selected as the first reference block (46-2), and blocks that match this reference block at 100% or 70-100% are assigned Hj = 0 and Hj = 1, respectively. An evaluation value is attached and the flag f of these blocks is set to “1” to indicate the completion of evaluation.
is set to This first evaluation loop (46-2 to 4
Among the song blocks whose evaluation was not determined in step 6-15), the youngest block becomes the reference block for the next evaluation loop (46-3, 46-4, 46-6), and the hierarchical structure of this reference block is The identifier is 2. Thereafter, the process is repeated in the same way. As a result, all blocks of the song are given a hierarchical structure identifier Hj.

The process shown in FIG. 41 is a process for converting the block-based hierarchical structure obtained in FIG. 40 into small-part hierarchical structure data. That is, the hierarchical structure identifier for the a-th measure is set in HIEa in FIG. 41.

Next, with reference to FIGS. 42 and 43, a description will be given of the process of extracting the tonality structure from the compound tone progression and setting the key of the scale for each section.

The idea underlying the processing is as follows.

(A) The tonality tends to stay the same as the song progresses rather than changing frequently.

(B) When the constituent tones of a compound tone include tones other than the scale tones of the previous key, the key should be transposed.

(C) When changing the key, it is desirable to change to a closely related □ key such as the F key or the subordinate key.

(D) For any key that has a total of 12 types of compound tones, when a note other than a scale note is included, the key in which the degree of coincidence between the compound note and the scale note is maximized is used.

In FIG. 42, initially set scale data (standard key, for example, scale data with C as the tonic) is set in SC 48-1. This SC is used as scale data for the immediately preceding section when determining the key of the current section.

At 48-2, a double note counter i is initialized, at 48-3, a counter C indicating the number of times the scale data has been inverted,
In 48-4, SC is substituted into a as the initial value for inverting m toward the major key (rotating up a perfect fifth) with respect to the so-called standard key, and As initialization for modulation, SC is assigned to b.

48-5, it is checked whether the i-th compound note is a subset of the scale a. For example, if the logical product of the compound note data CC and the scale data a matches the compound note data CC, then the compound note CC is checked. is a subset of scale a. When this condition is satisfied, a is assigned to SC and a is used as the scale data 5CALE i of the current section (48
-13), similarly, in 4g-6, it is checked whether the i-th compound note is a subset of the scale S, and if it is true, 48-
In step 7, assign b to SC and 5CALE i. If neither of these conditions holds true, turn a a perfect fifth up and turn it into 4.
8-7), invert b a perfect fifth down (4B-8), increment the inversion counter C (48-9), and repeat the key determination.

There are 12 types of keys in total, from 48-5 to 48-10.
The loop only needs to be repeated 6 times, i.e. 48-10
C>6 holds true when the complex tones in the current section include notes that are not included in the scale of any key.

In this case, the process 48-11, that is, the process shown in FIG. 43 is executed. In this process, a scale with the maximum degree of coincidence between the scale tones and the constituent tones of the compound tone is determined among the 12 scales.

That is, for each key of j = 1 to 12, calculate the degree of coincidence V between the scale with j as the key and the constituent notes of the compound note49-3), and set the maximum degree of coincidence at max, The key j that gives the maximum degree of agreement is determined as P (49-
6).Finally, the scale data obtained by rotating the scale data SC of the previous section by P rotations is adopted as the scale data 5CALEi of the current section, and this scale data is set to SC to determine the key of the next section. There is (49-9).

When the key of the current section is determined (48-13, 48-1
4, 48-11), the double note counter i is incremented 48-12), and the key determination process is repeated until the completion of all f double note progression data (4B-15).

For example, assume that the diatonic scale (do-re-mi-fa-solacido) is selected as the type of scale, and the key of the previous section is C, and the key of the current section is F. In this case, SC contains data representing (C, D, E, F, G, A, B). Therefore, in Figure 42, after the processing at 48-3 and 48-4, it is confirmed at 48-5 that the compound note is a subset of scale a (which is the same as SC), and it is confirmed that the compound tone is a subset of scale a (which is the same as SC), and it is The diatonic scale in key C, which is the same as the previous section, is used as the scale for this section. As can be seen from this example, when all of the constituent tones of the current section are on the scale of the key of the previous section, the key of the current section is maintained in the same key as the previous section.

As mentioned above as a composite of the current section, when D and F# are used instead of F, in 48-5 after 48-3.4, a is a diatonic scale with the key of C, so the compound note is the part of a. It is not a set, but after turning this a 5 degrees up in 48-7 to obtain the diatonic scale of G, when checking 48-5, the compound note becomes a subset of a. Then, the diatonic scale of G becomes the scale of the current interval. As can be seen from this example, when the constituent tones of the compound tone in the current section include notes that are not in the scale of the key of the previous section,
If we take the key of the previous section as the standard and repeat the modulation from this standard key to closely related keys (for example, C4G+D-+A・
), the first key of the scale that includes all the constituent tones of the compound tone is adopted as the key of the current section.

As will be described later, the scale data 5CALEi of each section determines the pitch structure of the melody used in the corresponding section. That is, the melody tone of each generation section is scale data 5CAL.
It is selected from among the scale notes defined by Ei.

Melody generation> The device of this embodiment is given external data that serves as the basis for composing, and after internally analyzing and evaluating the basic data,
Let's move on to melody generation.

FIG. 44 illustrates a schematic flow of melody generation shown in 4-8 of the composition general flow (FIG. 2). In FIG. 44, HIEi is the hierarchical structure data for each multitone section extracted in the previous multitone progression evaluation, and as shown in 53-2 to 53-4, this hierarchical structure data HIEi is the placement type ( It is used to control the generation of LL). Details of this control will be described later. Furthermore, the hierarchical structure data is 53
It can also be used to control the sound range of the stationary type (LL) as shown in -5' and 53-6.

The layout of the first stage melody forms the basic framework of the melody's tone sequence, and the range of the first stage melody basically regulates the range of the melody. In this embodiment, the hierarchical structure obtained from the multitone progression is used for the control, and this is also one of the features of the embodiment. However, it is not necessary to limit the control factors for the placement of the first stage melody only to the information on the hierarchical structure extracted from the compound note progression. For example, the hierarchical structure and random numbers are each weighted, and the The layout of the step melody may be controlled and the weight may be specified by the user.In short, the layout can be modified so that the user's compositional intentions are reflected in the generation of the layout.

The placement type (LL) of the first stage melody is multitone progression data CC
By using i, it is converted into a pitch expression format, that is, a melody data format (first stage melody) (53-
7), and the second stage melody sound is added to this first stage melody according to the production rules (
53-8), the production rules used in the addition of this second-stage melody sound are the same as those used to classify the second-stage melody sounds included in the motif, and therefore are reversible in terms of melody analysis and synthesis. Gender exists.

By adding the second stage melody to the first stage melody, the pitch sequence of the melody is completed. After the melody pitch sequence is completed, a melody note length sequence (rhythm pattern) is generated (53-9).

Here, we will introduce a basic rhythm pattern consisting of a predetermined number of notes (
The note length sequence determined in essence generation 4-4 or essence extraction 4-6) is transformed according to the pulse scale selected in initial setting 4-1, and converted into a note length sequence with the same number of notes as the pitch sequence of the melody. be done.

Generation of the first stage melody, addition of the second stage melody, and generation of tone length data are executed for each double tone section in the case of the flow shown in FIG. 44. Therefore, in 53-10, the melody data generated in a certain section is moved to a continuous area.

FIG. 45 is a detailed flowchart of generation, saving, and loading of the first stage melody type (53-5 in FIG. 44).
Details of 2.53-3.53-4) 0 In this example, control of the stationary LL using the hierarchical structure data HIEi is performed as follows. First, it is determined whether the passage of interest is structurally different from the previous passage by comparing the hierarchical structure data of the passage of interest with the past hierarchical structure data. A new LL is generated only for passages recognized as sections with different structures. This generation is performed based on the above-mentioned fixed type feature parameter PC.

For passages that are recognized to have a similar structure, a new positional type LL is not generated.Instead, among the previously generated positional types, a positional type generated in an interval with a similar structure to the section of interest is used. do.

For example, let's assume a piece of music consisting of four passages, and assume that the structures of these four passages are a, b, C, and a, respectively. Since the structure a of the first passage is a structure that has not been seen in the past, a placement pattern is generated according to the feature parameters of the placement pattern. Similarly, the second and third passages have new structures that have never been seen before, so the placement patterns are generated independently. However, the final passage has the same structure as the first passage of the song.
It is. Therefore, as the placement pattern for the final passage, the placement pattern generated for the first passage is used as is.

Generating a new placement pattern for a passage with a new structure means that a different motif is generated from the passage with this new structure. Now, if we were to use the placement pattern generated for the first measure of a passage repeatedly in the following measures of that passage, we would be aware of a motif that is one measure long. Generally, the motif length is 1 to 1 measure. It is several measures and 1
The length of the motif often changes during the song, the fourth
Taking this into account in the example of Figure 5, when a passage with a new structure is detected, l or 2 bars can be used as the length of the motif in that passage. When a two-bar motif is selected, placement patterns are generated independently for the first and second measures of the passage, with subsequent odd-numbered measures using the placement pattern of the first measure, and subsequent even-numbered measures using the placement pattern of the first measure. The measure uses the placement type of the second measure.

Reference to hierarchically structured data in past intervals.

To repeat the position type in the past section, place type (LL
) A data buffer is prepared. An example of a free-standing data buffer is shown in FIG.

To explain according to Fig. 45, passage 54-1 (39th
The bar counter in the barno section shown in the figure is set to “l”.
”.In 54-2, it is checked whether or not it is the start of a passage by comparing the hierarchical structure data HIEi of the current measure with the hierarchical structure data HIEi-1 of the immediately preceding measure.0For example, l HI E i HI E+1
The beginning of a passage is when l≧2 holds true. When it is determined that it is the start of a passage, the measure counter in the music place is reset to "l" (54-3), and then the stationary type data buffer is searched to determine whether the current passage is a passage for which a new stationary type should be generated. (54-4), the search of the fixed type data buffer is performed as follows. First, read the data at address 0 of the 9M data buffer (data indicating how many sets of position types have already been generated), and read the data (for each position type) whose addresses are the data indicated from the following addresses l to N. header information) and compares the hierarchical structure data shown in the upper 8 bits with the hierarchical structure data HIEi of the current measure. Hierarchical structure data similar to the hierarchical structure data HIEi of the current measure (for example, the same value as HIEi or (HIE r
If there is no data with a value of 1), the current measure is the first measure of the passage for which a new placement type is to be generated. When a header with similar hierarchical structure data is found, the following position-type data is loaded as the position-type of the current measure. In the case of a new passage, first it is determined how many measures the motif should consist of (54-5). This determination can be realized, for example, by random number generation. When it becomes a 2-bar motif, the flag fl is set to "1" (54-7, 54-8), and the placement type is newly set for the next bar m (the second bar m of the passage). Allow it to be generated. Then, the 47th part should generate a set type.
(see figure), save to a free-standing data buffer (54-9
), in detail, create a header using HIEiXO100 + measure counter value X0OIO + number of measures of the motif,
Increment the number N of sets of fixed types at address 0 of the fixed type data bar 2, write the address of the header to address N, and write the header, LLNO(
length of the generated stationary mold), LL], LL2...L
Write LLLNO (generated positional data), then perform other melody generation processing (54-10
), increment the bar counter (54-11)
.

If it is recognized in 54-2 that the passage has not changed, check flag fl (54-13), f
When sentence = 1, the current measure is the second measure of a passage that generates a 2-measure motif, so generate a fixed type again and load it into the fixed type data buffer 54-15), and set the flag fJL.
is reset to "0" (54-16). When f statement = 0, the header corresponding to the hierarchical structure data HIEi of the current measure is searched from the buffer, and the length of the motif indicated by the header is searched. When the information is 1 measure, the following position data is loaded into the header, and when the motif length is 2 measures, the remainder of 2 of the measure counter is compared with the measure number of the header, and if they match, the header is loaded. Loads the following position type data as the position type of the current measure.

Details of the generation of the first stage melody placement type executed at 54-9 and 54-15 in FIG. 45 are shown in FIG. 47, and 56-
The ckno of 1 represents the number of constituent tones of a complex tone. The number of constituent tones of a complex tone is obtained by counting the number of "l" among the 16 bits of the complex tone data CCi. In the example shown in Fig. 47, erl, which is a fixed type control parameter for generating PC, ~PCs, takes a random value in the range of 1-ckno and means the note number of a compound note (56-4), and r2 is PC3. The random value of the octave compound tones from (the lowest note of the stationary type) to PC2 (the highest note of the stationary type) is taken and represents the octave number of the LL to be generated (56-5). a=r+ +r2 X0
56-
7), When this candidate a satisfies the condition of pc, candidate a is adopted as LL (56-8.56-12.5
6-14), however, depending on the value of PC, the preceding L
After L (for example, the first LL) is determined, subsequent LL2 candidates may not satisfy the PC condition forever. For example, PC2 = 501 (the highest note of the overturn is the fifth
The first compound note of the octave), PC3=401 The lowest note of the C type is the first compound note of the fourth octave), PC
, = 3 (the maximum value of the difference between adjacent LLs is three tones), PCs = 3 (the minimum value of the difference between adjacent LLs is three tones), and LLI = 403.
(If the first LL is the first compound tone of the fourth octave), then to match the conditions of PCa and PCs, LL2 = 503 or 303 (when the number of constituent notes is 3)
This does not meet the conditions of PC2 and PC3. To prevent this, prepare a loop counter LOOPC, and set the loop counter LOOPC to a certain value (for example, 100).
If the result is above, candidate a is forcibly adopted as LL (56-9.56-10.56-11).

FIG. 48 shows details of check 56-8 in FIG. 47.

For candidate a of LLi to satisfy the PC condition, (a) a≦PC2 (be below the highest note) (b) a≧
PC3 (Must be at least the lowest note) (c) l a
LLI -1l≦PCa (The difference from the previous (7) LL is less than or equal to the maximum value PCa) (d)' I a L Lt 11≧PCs
(The difference from LL (immediately before) is greater than or equal to the minimum value PCs) must hold true.In the flow of Figure 48, when these conditions do not hold, the flag OK is set to "0" (Figure 48). Olda in the middle represents the previous LL, 5
6-13).

FIG. 49 shows details of 53-7 in FIG. 44.

The purpose is to convert the format of the LL of the first stage melody indicated by (octave number 10 compound note numbers) into compound note data C.
C to convert it into the format of melody pitch data indicated by (octave number ten syllable name number) and store it in MEDi. 58-5. The processing in 58-6 is performed when the number of constituent tones of LLi (LUiΔ0Off) is larger than the number of constituent tones (CKNO) of the complex tones in the current section.
In the figure, C is the counter of the constituent tones of the complex tone, LLiAffOO is the octave number of LLi, and j is the octave number of LLi. This is a note name counter.

Figures 50 and 51 are the second part in 53-8 of Figure 53.
This is the details of adding step melody sounds. The purpose of this processing is to complete the pitch sequence of the melody by adding desired second-stage melody sounds to the first-stage melody. In order to add the second stage melody sounds, we need the characteristics of the second stage melody mentioned above (R5i), scale data with tonality (SCALE i) obtained from the multitone progression evaluation, and knowledge for classifying the second stage melody sounds. Production rules to express are used. The added second stage melody sound must satisfy the following conditions.

(B) The sound must be within the specified range (B) The scale note must be the one indicated by the scale data obtained from the compound tone progression evaluation (C) The note must not be a component of a compound tone (D) The conclusion and plan obtained from the production rules In FIG. 50, the outer loops 59-4 to 59-18 are loops that repeat as many times as the planned number of second-stage melody sound identifiers R3i, and 59 -5 to 59-16 loops are repeated as many times as there are notes in the first stage melody, 59-
8 to 59-14, the lower limit n is used as a candidate for the second stage melody sound.

Each pitch data k within the range from up to the upper limit up is sequentially examined (see Figure 61). 9),
Calculate the function F (59-10), perform forward inference using the production rule (59-11), and check whether the conclusion matches the planned second stage melody note identifier R5i (59-11). ), when they match,
The pitch data satisfies all of the conditions for the second stage melody sound mentioned above. Therefore, the second
Increment the counter nct Ct that counts the number of stage melody sounds, enter it into the pitch data kQVMnctct of the found second stage melody sound, set the additional position 1j of the second stage melody sound in PO3Tnctct, and set the related flag fij set to “l” (59-19
~59-22) 0 In this example, at most one second stage melody sound is added between the first stage melody sounds, and ffLj = 0 means that the first stage melody sound MEDj
,! = Indicates that the second stage melody sound has not been added yet during MED4.1.

If the condition of conclusion=R5i in 59-12 is not satisfied, the pitch data of interest does not satisfy the condition as a second stage melody sound.

The pitch data k is incremented (59-13) and the process is repeated. When k>UP is established in 59-14, the first stage melody tones MEDj and MED4-+
This means that the second stage melody sound was not added during this period. Therefore, by incrementing j, (59-1
5) Proceed to a test to see if a second stage melody sound can be added between the next first stage melody sounds.

Details of the setting of candidate sounds in step 59-6 are shown in FIG. In this example, the first stage melody sound around 1 M E
The pitch range to be searched is from 5 semitones above the higher note of D i and MEDI4+ to 5 semitones below the lower note (62-5 to 62-7), where i = 0 , that is, when trying to add the second stage melody sound before the first i1 stage melody sound of the section of interest,
The pitch range is from one semitone above to five semitones below 62-1.2)
, when i = EDNO, that is, when trying to add the second stage melody sound after the last first stage melody sound of the section of interest, add the second stage melody sound 5 to 5 semitones above the last first stage melody sound. The pitch range is up to a semitone below (
62-3.4).

The details of the calculation of F in 59-10 are shown in FIG.
In this example, only one second-stage melody sound is added between the preceding and succeeding first-stage melodies, so some functions (Fl to F3 in the figure) are set to predetermined values. .

Details of the forward inference of 59-11 are shown in FIG. 44 mentioned above.

When the processing in FIG. 50 is completed, the total number of added second-stage melody sounds is stored in nctct, and the array (V
i), the pitch data of the second stage melody tone added to the i-th in the process of FIG. 59 is stored, and the i-th of the array (PO5Ti) is stored in the i-th in the process of FIG. The position information of the i-th added second stage melody sound is stored.

These data are converted into a melody pitch sequence (VMEDi) format by the process shown in FIG. Note that the array (VMEDi) is initially set to arpeggio (MEDi). 60-2 to 60-9 are in order of addition position,
This is a process of sorting the arrays (PO5Ti) and (VMi). 1 position data PO from 60-10 to 60-19
The pitch data VMi of the second stage melody tone is inserted at the position indicated by 5Ti.

In this example, only one second stage melody sound can be added between the first stage melody sounds, but the process may be changed so that a plurality of second stage melody sounds can be added.

At this point, the pitch sequence of the melody is complete. The remaining processing is to generate a string of note lengths for the melody.

Figure 55 shows the flow of generating a melody note length sequence (5
(Details of 3-9) First, the number of notes of the standard rhythm pattern obtained by essence generation or extraction is the number of notes generated in the section in focus. 7 layers edn.

(the number of data in the pitch sequence of the melody) and calculate the difference a between the two (6B-1). When the number of generated notes is less than the number of notes in the reference rhythm pattern (when a>0) )
The optimal combination of notes using a pulse scale is repeatedly executed by the number of differences (66-2 to 66-6), and when the number of generated notes is greater than the number of notes in the standard rhythm pattern (a
go), the optimum division of notes is repeatedly executed by the difference (66-7 to 66-11). In this example, 16-bit data is used as the data format of the rhythm pattern, and each bit position is Assign “l” to each timing.
Since this indicates that a sound is generated at a bit position with a value of
2).

The details of the joining process are shown in Figure 56.
j represents the j-th component scale of the pulse scale to be used, and RR is the rhythm pattern to be processed. R.R.
For example, when the standard rhythm pattern is ノ `` - こ ノ j , the normal note is If you combine one note using the pulse scale (see Figure 7), the result is Tsunotsuno.

This can be obtained as follows.

First, RR is initially Al, while the normal pulse scale is 1213
1214 1213.1215. RR bit “
The point where the normal pulse scale has the smallest weight among 1'' is the 7th position from the right end.The bit at this position becomes "0", so the resulting RR is, which is It represents no no no no.

The details of the division process are shown in FIG. 57. 0 division is executed by setting the point where the weight of the pulse scale is maximum among the bits of RR as "0" to "l". For Noniko 〒】, when we divide the notes using the normal pulse scale, the result is.

nnn becomes.

The details of the conversion 86-12 to the MER data format are shown in Figure i58. In Figure 0, cl is a note counter and C? is a counter that measures the length of each note. In this example, ME
Since Ro contains the length from RR until the first "1" appears, notes that cross the boundary line (bar line) of the section are also processed parts (to prevent syncopation).

The details of the data movement 53-10 are shown in FIG. 59. First,
MERo (blank part at the beginning of the current generated section) is added to the generated note length data MELR+5eldno of the last note. Here meldno represents the number of notes that have already been generated, the pitch sequence VMEDI ~ VMED generated this time
Move vmedno to MELD and generate the tone length sequence M ER+ ~M ER vsedn.

is moved to MELR (70-2 to 7O-6).

meldno is updated and the process ends (70-7).

Summary of the Music Composer> As is clear from the above description, this device as an automatic music composer has various features, some of which are listed below.

(A) Progress information of multiple tones (primary constituent tones) is input in order to define a set of pitches of the first stage melody forming the skeleton of the melody. (B) The pitch sequence of the melody is generated by a two-stage process in which a melody is generated in the first stage and then a melody in the second stage is added.

(C) It incorporates a production system that uses musical knowledge to make inferences in melody analysis and synthesis.

(D) Since the process of classifying the second-stage melody sounds included in the melody and the process of adding the second-stage melody to the first-stage melody are performed based on production rule data that expresses the same musical knowledge, Both processes are reversible.

(E) A song is planned by analyzing the multitone progression given as material for composition and extracting the hierarchical structure and tonality structure of the song.

(F) The extracted tonality structure defines the key of the scale used in each section. This ensures a tonally controlled song.

CG') The extracted hierarchical structure is used to control the generation of the first stage melody. This allows for variety and consistency in the generated songs.

Note that the above-mentioned embodiment is merely an example, and various modifications, changes, and improvements are possible.
Although it is used as control data in the generation of L, the stationary feature PC may also be changed as the song progresses. This can be achieved with

Similarly, the characteristics of the second stage melody (R3i) can be changed as the song progresses6. For example,
One second-stage melody sound identifier among the characteristics of the second-stage melody extracted from the motif is replaced with another second-stage melody sound identifier. This can be achieved by randomly selecting one from a set of second-stage melody tone identifiers.

Regarding the rhythm, in the above embodiment, the rhythm is controlled by combining notes using a pulse scale into one division, but the dominant mini-rhythm pattern included in the motif is extracted, and the tone length of the melody to generate this pattern is controlled. It may also be included in a column.

[Effects of the Invention] As described above, the automatic composing machine of the present invention uses, as basic data for composing, the motif and the pitch used as the sound of the first stage melody to be included in the melody line to be generated. The type (primary constituent sound) progression information is input from the input means, and the feature parameter extraction means analyzes the motif according to the given progression information of the primary constituent sound.
The characteristic parameters of the melody in each stage and the characteristic parameters of the second stage melody are extracted, and the characteristic parameters of the melody in the first stage are followed by the first stage melody generation means using the data of the primary constituent sounds in each melody generation section. A first stage melody is generated, and a second stage melody according to the second stage melody characteristic parameter is added to the first stage melody by a second stage melody adding means. Therefore, the melodies of the first stage form the framework of the final melody, and the melodies of the second stage serve to decorate, connect, and support this framework.

This automatic composing machine generates a melody by interpreting a given motif from the perspective of a multitone progression, and can plan different pieces of music depending on what kind of progression is given as a multitone progression. Therefore, it is possible to compose a wide variety of songs, from harmonic to non-harmonic. Furthermore, since there is no need to input chord progressions, the user does not need to have any knowledge of harmony.

[Brief explanation of the drawing]

FIG. 1 is an overall configuration diagram of an automatic music composer according to an embodiment of the present invention, FIG. 2 is a flowchart showing the overall operation in music composer mode, and FIG. 3 is a diagram of the main variables used in the music composition process. Diagram showing the list, Figure 4, Figure 5, ft56
Figures 7 and 8 are diagrams showing the format of the data used, Figure 9 is a flowchart of initial settings, and Figure 1θ is an example of compound tone progression data stored in the compound tone (primary constituent tone) progression memory. , FIG. 11 is a flowchart for reading multitone progression data, FIG. 12 is a diagram illustrating pulse scale data stored in the pulse scale memory, FIG. 13 is a flowchart for reading pulse scale data,
FIG. 14 is a diagram illustrating production rule data stored in the production rule data memory, and FIG.
Figure 16 is a flowchart for reading production rule data, Figure 16 is a diagram illustrating melody data (motif data) stored in the motif memory, Figure 17 is a flowchart for loading melody data, and Figure 18 is a flowchart for essence generation. , FIG. 19 is a flowchart for setting the characteristics of the melody sound type in the first stage,
Figure 0 is a flowchart for setting the characteristics and characteristics of the second stage melody, Figure 21 is a flowchart for evaluating the motif rhythm by section, and Figure 22 is a flowchart for setting the characteristics and characteristics of the second stage melody.
Fig. 23 is a flowchart for calculating Ps and Pss; Fig. 24 is a flowchart for calculating re and Fee; Fig. 25 is a flowchart for extracting the first stage melody note type of the motif; Fig. 26 is a flowchart for extracting the characteristics of the melody sound type in the first stage,
Figure 7 is a flowchart for extracting the characteristics of the second stage melody, Figure 28 is a two-row chart for classifying melody sounds into first stage melody sounds and second stage melody sounds based on compound note data, and Figure 29 is a flowchart for extracting the characteristics of the second stage melody. A flowchart for calculating the function F that represents the state of the melody, Fig. 30 is the function F
31 is a flowchart for calculating function F2, FIG. 32 is a flowchart for calculating function F3, FIG. 33 is a flowchart for calculating function F4, FIG. 34 is a flowchart for calculating function F5, 35th
The figure is a flowchart for calculating function F6, Figure 36 is a flowchart for calculating function F7 and Fs, Figure 37 is a flowchart for temporarily storing the calculated function F, and Figure 38 is for inferring the type of second stage melody sound. flowchart,
Figure 39 is a flowchart for calculating the degree of coincidence of compound note progressions between blocks, Figure 40 is a flowchart for generating hierarchical structure data from the calculated degree of coincidence, and Figure 41 is a flowchart for calculating the degree of coincidence of compound note progressions between blocks. Flowchart for converting into structural data, Figures 42 and 43 are flowcharts for determining the key and scale of each section from a compound note progression, Figure 44 is a flowchart for melody generation, and Figure 45
The figure is a flowchart showing the generation, saving, and loading of the first-stage melody sound type, Figure 46 is a diagram illustrating a fixed-type data buffer, Figure 47 is a flowchart of sound type generation, and Figure 48 is a check in sound type generation. flow chart,
Figure 49 is a flowchart for converting a fixed type into melody data format, Figures 50 and 51 are flowcharts for adding second stage melody sounds to first stage melody, and Figure 52 is a flowchart for adding second stage melody sound materials and processing. Figure 5 showing the order of
Figure 3 is a flowchart for setting the range of pitches that are candidates for the second stage melody sound, Figure 54 is a flowchart for calculating function F, Figure 55 is a flowchart for generating melody note length data, and Figure 56. is a flowchart for optimal combination of notes, Figure 57 is a flowchart for optimal division of notes, Figure 58 is a flowchart for converting the generated rhythm pattern into MER data format, and Figure 59 is for moving generated melody data to a continuous area. It is a flowchart. l...CPU, 2...Input bag δ, 3.
...Motif memory, 4...Multiple note progression memory, CCi...Multiple note data, LLi...
...First stage melody sound type, PSi...Features of second stage melody, PCi...Characteristics of first stage melody sound type. Patent applicant: Casio Computer Co., Ltd. IL reader 5CALEi
iiwγ-daP! L 5 Lis Lis Melo Ten〉 Fig. 4 Deda Hyouki 1) Ku Merote ni Kiki Hataiho l Lame-8〉 Sai 11 Melwo'4 LLi
0402PC3 (LLiII') Poegum R51 (! Dormitory 1 Saki 11 Meloten's feathers - number) (I? Io
cl, I, No. 1711 0101 (Sex) Figure 5 Goji Karasuto <'Ji 1 fi'tm D7-J'! M9IL)
? 5 Me 4 > Figure 6 data added J3] PSCALEi (re) L-Susun 7-) L) Sumi 11) 1, Normano L 1111 1111 1111
11112, η J': 1111 1111
1111 1111 Fig. 7 1Σ゛-ノ? Wohishiki J (4) Fig. 16 Merote°(U”-〃Ban 1) ii: C, E, G MD6 Tomi OS OOMRs Seal 4 CC4-0091CRt-10 Fig. 8 °-Tata sweaty Kinuta row Figure 9 t77 Handling Figure 10 Ne! ,1111,01
01,0001 (Normal) (With 5I) ノV几susuno-9 te ゛-j 11 Figure 13 ψ ノL sufunofuri L te 1 noa 0 bite down--chi 3≧,-
treason) "l"j" (-L x -U
Y N 7 do c
'LI Q2023 02
Q 8 0 -1 -2
53 Q 1 0 4
5 104 0 7
0 -3 -4 1s5
0 t, 0 9 6
207 3 B @-
-2B 309 -as 8
-3 -3 -6 3510
1 ' 5 ' 1 -7
It, 4511 4
5 4 -7 -8 S. Figure 14 Arodaku Himself Yoshiru 几zartai string program η Jo 5 no L-ru O Guo

[Include δ Figure 17 Riding the Norote Rider Figure 18 Ella 1-5 again) i3 Father No. 191! ! Suni 1 Kouji Hatake Melo Ia at:r, Shitetsu 20th Figure 2 Toge” 1 Me 0 Ry 1 Kiben 6 Han Risu ° Otsue〒Iiro 1・ IJJI b=-2 1, I JIJ, II ””
(A-1jlJξJbar-2 Fig. 22 Ps, Pe, Pss, Pee) 'jlL'
3rd floor snow 0 ~ Exit 3 8:00-〇Pss2 P2H4 Pe-3Pee-4 (J ON: N Figure 23 Ps, Pss (Osu1 Tomoe Figure 24 Pe, Pee 6') Usude Figure 26 1st blow, 1st round, 1st class, 1st class, 1st grade, J official, YxV, Hayabusa, 2IJ, 1st class, ≠ 1st class.
0 Figure Fl: Large 1 ξ 1 Hansui Merote 1 Cloth 1] I-cho Sonozuki City No. 3
1 Figure F2-1iAf)'41PtN:An"f4Ts12f
laM'b' Figure 32 F3: @Later night 1jii Haru Nurote 1 also nine lengths 01 eyebrows length F41
1 illness built 1
Figure 37: F1 to Fn (n-8) - Kamome 1 Figure 38 Maegashiron Figure 41: 1 Kaya, data to Tomoe 42nd scale Su-Ten-Nine-handed criminal n17'T-JLIPIl'F'yt court [l)N
Gun N (AN) ☆ to the key? 11 F F F F (Hex) Fig. 46 right "1L (LL) u° - Fufu 7 Fig. 48 ÷ Esoku Fig. 52 e%040a 40 b 040o 4.0502 P" - g Written amendment (method) % formula % 1. Indication of the case 1985 Patent Application No. 325180 Address 2-6-1 Nishi-Shinjuku, Shinjuku-ku, Tokyo Name (14
4) Tadashi Kashio, Representative of Casio Computer Co., Ltd.
Male 4, agent address: T-S Building 3, 1-13-4 Nishi-Shinbashi, Minato-ku, Tokyo
Floor, 5'2' -- Name Patent Attorney (77, 34) Town 1) Masa Toshi. -1 Co. 5. Date of notice of reasons for refusal: March 29, 1985 (starting gate) 6. Drawings subject to amendment. 7. Contents of corrections Figures 1 to 9, 11-41, and 43 to 59 of the drawings will be corrected as shown in the attached sheet. 1st figure & body a ~ 4 songs v3 floating float 11 Ei 11 police tax data 5 CALEi
1 Tsukasa 1 data list
[Kusui Zloten's confirmation] Meeting D10 + (Hex) Ku Irochi °in 1 trillion γ-ter - 16 bit growl" Momme lower 8 bits is the chief tader - 1 chief's Somoto Rin's Tachicho 0! 4 cans in 1 ma
Urawa (Fig. 4 Koshi-Yu 1 Risoq (1) 0124Q Aigawa) 1) 1 give 3 Snow day = q Kusu-Tai-no L (doremi) so'' 5% 7
K ) Li-5FxilUi Figure 6 Type yr4 formula (3) PSCALEi (lY7L'; zh-Jt) (1v
11) 1, Numaru 1111 1111 1111
tilloooo ooooi oooo
ooooooooo 0000 0000 0001
2.9 Shiba' till 1111 111
1 1111 Figure 7 - Typical knowledge (4) No 1! +tv 1#WL
MDi MRi
KO Vless IC4" 040020 2D4" 040222 3 E4J'0404 2 t. 4F4111) o4o526 5 G4-0407 4
86 csJosoo
4 10MC1+Tomi 0400
MR+=2 (716 way to t) MD2-
0402 MR2-2MD3=0404
MR3= 2MD4 Snow 0405
MR4=2MDs=0407
MRs-4MD6e0500 MR
6= 4CC+Snow 0000 CR+=I
OLL+Snow 0401 p(1-4LL2
=0402 PC2=0501LL3
=0403 PC3=0401LL4
=0501 PC4=1 te°°-Yuanshiki゛子11 +hf)ri(i) CCi
CRi C
Ri VVless (16” (16m) (161L)
(10 blood)'Trless take 10 5, FFFF, 5555,1111,010
1,000? Fig. 12 Wiss Scanner 11 Fig. 13 Bar IL Susmo Arr 79f > lb publication Mui Rule >t<-L X U
Y N Finished dress Figure 14 70 Rikujo) 几-lute°=Tie 2f11 Figure 15 Frota η John Lu JLo theory 4 and Merotete °-91 Tokomi Sei'1lffx Lochi'4 'fj A'I Okishi vertical 6 ! Second
0 Figure 1 PzaPm! r）mtW Rism katsu〒I0 Ps, PePssPeeri Kaoru ratio No. 24 Fig. Fe, Pee') II lt3 Fig. 26 Completion of L #shi big l matter'/QMF operation No. 28 Design 1 #ii floor Melothea Ken/92 Atrocious Melof λ1's version 3-
1F1: πf) Jaw voice old 1f number first ＼ Fig. 36 F7: I death direction harmony 1y-7 group education Kaya Takashi F8: JI
I+'<1;
and Fig. 37 F1 to Fn (n-8) - day before Tanezono reasoning Fig. 43 Lesson '7-/L hand + 1 constant jj amount (2) After the system゛-evening (1 shot) O what? Ku9pu!
(7)) (Group N) 1 Wolf 1's own 1L category 7 dress (A1) 2
$2 (A2)N Our company N
(AN) A1 Heshita 1 A1 Ya 11 type Ω series 4 chestnuts are also "tsu, (L L) tegu 8 tsufa ckno-4 48th figure 52 (end) MEDW4T thick 0 040a 40 b oso. 040a 4-0502 Fig. 53 41 Gift of the gift Fig. 54 Ff) % - thin MERff - evening WS %'E change center

Claims (1)

[Scope of Claims] Motif input means for inputting a motif; primary constituent note progression input means for inputting a progression of primary constituent notes that defines the notes to be included in the first stage melody in a melody line; and a section of the motif. feature parameter extraction means for analyzing the motif and extracting characteristic parameters of the first stage melody and the second stage melody according to the progression of the primary constituent sounds in each melody generation section; 1
a first stage melody generating means for generating a first stage melody according to the characteristic parameters of the stage melody; An automatic music composition machine comprising: second stage melody adding means for adding a stage melody.