JP2526830B2 - Automatic chord adding device - Google Patents

Description

TECHNICAL FIELD OF THE INVENTION The present invention relates to an automatic chord adding device (hereinafter, automatic chord) capable of automatically adding a chord (hereinafter, referred to as chord or chord) to a given melody. Addition device or automatic code addition device).

[Prior Art] An automatic code adding device for automatically adding a code to a given melody is already known.

For example, the automatic code adding device disclosed in JP-A-58-87593 determines a code as follows.

(B) For the 12 tones, calculate Σ (note scale x note value) -Σ (outside note x note value) and select the maximum note.

(B) The final note of the melody is used to determine whether the key is a major key or a minor key (note that (a) does not distinguish parallel key, for example, c major and a minor key).

(C) Select seven chords (diatonic chords) from the keys determined in (A) and (B) above (for example, in C major, Cmaj, Dmin, Emin, Fmaj, Gmaj, Am
Select in, Bdim).

(D) For each block (measure), the compatibility (similarity) between the constituent notes of each of the seven chords and the pitch of the melody is calculated.

(E) The fitness is multiplied by the priority according to the chord priority table (transition frequency of two adjacent chords).

(F) Of the seven types of codes, the code that maximizes the evaluation value = the goodness of fit × the priority is determined.

The applicant of the present application also discloses an automatic code adding device in Japanese Patent Laid-Open No. 58-114097. This automatic code adding device adds a code as follows.

(A) Tonality is determined by the final tone.

(B) The dominant sound (tongue) is determined by the melody in the block (measure).

(C) The chord is determined by referring to the table based on the tonic and the previous chord.

However, in each of the conventional examples, there is an implicit limiting condition for the melody that is input (the melody must be input until the end of the song. Do not transpose in the middle of the song). In addition, since there is only one chord progression as an output, other influential chord progressions are discarded. There is a problem that the same result is obtained even if the order of the sounds of the melody is changed.

[Object of the Invention] Therefore, an object of the present invention is to provide an automatic chord adding device that adds chords based on a more musical basis without limiting conditions for input melody and without discarding chords that can be added. Is to provide.

[Configuration of the Invention] In order to achieve the above object, the present invention provides an automatic chord adding device that automatically adds chords to a given melody in a time series of sounds that compose the melody (hereinafter,
A melody input means for inputting data representing a melody tone sequence), a chord candidate assigning means for assigning chord candidates to a predetermined section of the melody, and a non-harmonic included in the melody according to the chord and the melody tone sequence. Knowledge storage means for storing musical knowledge for classifying voices, and applying the music knowledge to the melody and chord candidates in the predetermined section, that is, the chord candidates to the melody in the predetermined section. Assuming that the chord is a chord, an inference means for performing classification of non-harmonic sounds on a melody sound that is not a constituent sound of this chord, and if the classification result of the inference means is a predetermined classification result, the chord candidate is set to the predetermined chord. It is determined that the chord is a chord with respect to the melody of the section, and if the result is not the predetermined classification result, it is determined that the chord candidate is not the chord with respect to the melody of the predetermined section. To provide an automatic chord adding device, characterized in that it comprises a means.

[Operation of the Invention] It is assumed that a melody in a predetermined section, that is, a melody to be evaluated and chord candidates, that is, chord candidates are given. On the other hand, the inference means applies the knowledge stored in the knowledge storage means to perform the classification of non-harmonic sounds. At that time, the inference means assumes that the melody sound that is the same as the constituent sound of the chord candidate is a harmony sound. If this harmony sound assumption is correct, a given classification result is obtained as a result of the application of knowledge. That is, the remaining melody tones are concluded as respective non-harmonic tones (because the melody tones are composed of harmony and non-harmonic tones). However, if the assumption of harmony is not correct, the above classification result cannot be obtained when knowledge is applied, and it can be concluded that a certain melody sound is a sound that does not correspond to any non-harmonic sound. . This is due to the first assumption, namely that the code candidate was not valid. Therefore, in the present invention, when such a classification result is obtained, the determination means determines that the chord candidate cannot be added to the melody. That is,
It is determined that the chord does not correspond to the melody in the predetermined section.

As described above, according to the present invention, chord candidates that are musically given a reason by using the knowledge stored in the knowledge storage means (which notes of a melody are harmony sounds and which sounds are nonharmonic sounds) Only the chord candidates that can be explained are the chords that can be added to the melody. That is, only the chord candidates that have produced the predetermined classification result are set as chords (codes that can be added to the melody) for the melody in the predetermined section. Since the classification knowledge of non-harmonic sounds does not depend on a specific key, it is possible to add chords without applying a limiting condition to the melody input as in the conventional technique. Further, since the above knowledge considers the order (sound sequence) of the melody sounds, it is possible to interpret the code according to the characteristics of the melody line. Furthermore, when a plurality of chord progressions can be added, it is possible to output all of them.

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

<Overall Configuration> First, the overall configuration of the automatic code adding apparatus according to the present embodiment will be described.
Shown in the figure. Reference numeral 1 is a CPU, which performs overall control for code addition. The work memory 2 is used to temporarily store code data and the like when an operation such as adding a code is performed. The input device 3 is used to input a melody or the like. As the melody input device, for example, a keyboard (real time input) can be used. The monitor 4 is composed of a CRT, a tone forming circuit, a sound system, etc., and displays and outputs a melody and a chord (accompaniment). The melody data memory 5 stores the input melody data. The chord memory 6 stores a set of chords, and the chord component sound memory 7 stores data of the chord constituent sounds. The production rule data memory 8 stores data expressing music knowledge for classifying nonharmonic sounds. In operation,
The CPU 1 determines chords that can be added to the melody of each section (measure) in the melody data memory 5. That is, the assumed code is read from the code memory 6,
The knowledge read from the production rule data memory 8 is applied to the constituent sound data and the melody data. As a result, when an illegal sound is found, the assumed code is regarded as an error and the code is regarded as a non-addable code.

<Overall Operation> FIG. 2 shows a general flow of automatic chord addition to a melody.

First, in 2-1 the melody data is read from the melody data memory 5, and in 2-2, the bar counter bar is initialized to the first bar.

Then, as shown in 2-3, the melody data corresponding to each bar is selected. Further, production rule data expressing knowledge for classifying non-harmonic sounds is read (2-4). Next, the code counter CCC is initialized (2-
5) Read the candidate code from the code memory 6 (2
-6, 2-7). Next, the non-harmonic sound is classified based on the read melody data, chord data, and production rule data, and it is determined whether chords can be added (2-8). In this example, all chords are sequentially applied to the melody of the section of interest, and it is determined whether or not each chord can be added (2-6 to 2-9).

When the evaluation of whether or not the chord can be added to the melody for one measure is completed, the process proceeds to the next melody section (2-11), and the process is repeated until the evaluation for all measures is completed (2-10).

<Data Format> The data format used in the flow described later is shown in FIG.
In this example, the melody data is represented by a string of pitch data and pitch length data, and the pitch data is 16 bits long, the upper 8 bits represent an octave, and the lower 8 bits are C for each semitone. It is represented by an integer value that is incremented by 1. In the note length data, the shortest note length is represented by "1", and the doubled note length is represented by an integer value such as "2".

On the other hand, the code data is 16 bits long, and the upper 8 bits represent the type (code type) and the lower 8 bits represent the root note of the chord.

<Reading Melody Data> FIG. 5 shows the details of reading the melody data executed in 2-1 of the general flow of FIG. FIG. 4 shows an example of melody data stored in the melody data memory 5. As a result of the processing of FIG. 5, the i-th pitch data is MDi
, The i-th tone length data is stored in MRi, and the total number of melody tones is stored in MDNO.

<Selection of Melody Bar> Details of the selection of the measure of the melody executed in 2-3 of FIG. 2 are shown in FIGS. 6 to 8. By this process, Ps contains the number of the melody data {MDi}, {MRi} of the first note of the bar of interest, and Pe
Contains the order of the melody data, which is one note before the first note of the measure next to the measure of interest. The information of Ps and Pe is used for the determination process 2 described later.
8 defines the range of melody data to be investigated.

<Reading Production Rule Data> Details of reading the production rule data executed in 2-4 of the general flow are shown in FIG. Below the figure, an example of production rule data stored in the production rule data memory 8 is shown.

The entire production rule expresses musical knowledge (see FIG. 10) for classifying the non-harmonic sounds included in the melody, and each production rule data is
The lower limit data Li as the data that defines the premise of the rule,
Function instruction data Xi that indicates the type of function, upper limit data Ui
With the data Yi and Ni as the conclusion part of the rule.
The function is a numerical expression of the characteristics of the melody to be analyzed.
An example thereof is shown in FIG. 15 described later. The value Fxi of the function indicated by the data Xi is greater than or equal to Li and less than or equal to Ui (Li ≦ Fxi ≦
Ui) is the premise of a production rule (proposition)
Then, the conclusion when this premise is satisfied is shown by the data Yi, and the conclusion when this premise is not satisfied is shown by the data Ni. When the data Yi or Ni has a positive value, that value indicates the number of the production rule to be referred to next in the forward inference described later, and when it has a negative value, it is not summed by its absolute value. The type of voice sound is expressed. Forward inference always starts with one rule, and this rule is called a root. Forward finding ends when it finds a negative conclusion Yi or Ni. This negative value includes data representing an illegal sound (sound that does not correspond to any non-harmonic sound) indicated by ill in FIG. As will be described later, when such an illegal sound comes to the conclusion of inference, it is considered that there is an error due to the assumption of the chord that is the premise of inference, and it is determined that the chord cannot be added to the melody. .

In the case of the address assignment of the production rule data shown in FIG. 9, the data of each production rule is stored in five consecutive addresses starting from the address of the lower limit data Li. For details, Li is an address divisible by 5.
Xi is stored at the address of remainder 1 divided by 5, Ui is stored at the address of remainder 2, Yi is stored at the address of remainder 3, and Ni is stored at the address of remainder 4.

In FIG. 9, ruleno is set to the total number of production rules.

<Reading Code Data> Figure 11 shows the details of reading the code data executed in 2-7 of the general flow. An example of codes stored in the code memory 6 is shown in the upper part of FIG. In this example, the major chords are C, C ^#, ...
It is stored in the order of B, and the minor code from the address 12 is stored in the order of C, C ^# ... B.

The data read from the (code counter (CCC) -1) address is used as a code candidate (29-1).

<Judgment> Next, processing 2-8, which is the main part of the general flow.
(Process of classifying non-harmonic sounds of melody for chord candidates and determining whether or not chord candidates can be added) will be described in detail. The flow of processing 2-8 is shown in FIG.

First, in 30-1, the non-harmonic sound of the melody when the chord candidate is assumed is classified, and the register RS ₁
~ RS _RSNO stores the discrimination value of non-harmonic sound or illegal sound. A counter rsc for this identification value is initialized to "1" (30-2), and it is checked whether RS _rsc is a value representing an illegal sound (ill) (30-3). If any of RS ₁ to RS _RSNO contains an illegal sound, the code given by CCC shall be judged to be a use digit according to 30-7. Only when all of RS ₁ to RS _RSNO have classified non-harmonic discriminant values (30-3 to 30-5), it is judged that the code given by CCC is usable (30-6). The memory of this judgment is, for example, root CH (root) = (CCC-1) mod12 type (type) = (int) {(CCC-1) / 12} corresponding to the area CHORD _root , _type This is done by writing a value.

<Classification of Nonharmonic Sounds> FIG. 13 shows details of the classification of nonharmonic sounds executed in 30-1 of FIG.

First, 33-2 initializes the non-harmonic counter j,
At -3, the note counter i of the melody data is set to the head position Ps (value obtained in the process 2-3) of the bar of interest. In 33-4, it is judged whether or not the i-th note of interest is a harmony sound. Harmonic sounds are the same as the constituent sounds of the chord candidate. If it is not a harmony tone, the non-harmonic tone counter is incremented (33-5), and the function F representing the characteristic element of the melody centered on the note of interest is calculated (33
-6), the forward inference based on the production rule is executed (33-7), and the conclusion about the note of interest is substituted into RSj (33-8). The note counter i is incremented (33
-9) Until the last note of the bar of interest is reached (33-10), the processing from 33-4 is repeated.

Details of the process 33-4 are shown in FIG. In 34-1, the chord data is converted into chord component sound data. This conversion is performed by using the type data of the chord data as a pointer, from the chord component tone memory 7 to the component tone data represented in the C tone (for example, 00001001000 in the case of major).
It is performed by reading 1) and turning it left by the number indicated by the root note data of the chord data. 34-2, 34
The melody pitch data MDi is converted into data of the same format as the constituent sound data CC generated in -1. When the result of CC ∧ mm (logical product) is not 0, the same sound as the melody sound is included in the chord constituent sounds. That is, MDi is a harmony sound (34-3, 34-5). When CC∧mm = 0, MDi is a non-harmonic sound (34-3, 34-4).

Details of the calculation 33-6 of F in FIG. 13 are shown in FIG. In this example, the function F is as follows: F ₁ : how many notes (non-harmonic) the note of interest is before (harmonic position of the rear) F ₂ : how many notes before the note of interest Is located (the position of the front chord) F ₃ : The number of non-harmonics between the front and rear chords F ₄ : The pitch difference between the front and rear chords F ₅ : forward harmonic sound and rear harmonic tone height of nonharmonic tones distribution between the F _6: forward harmonic tone from whether high melody sound to the rear harmony sound is changed monotonically F _7: and the rear harmony sound Pitch difference with the preceding sound (pitch progression to the rear harmony) F ₈ : Pitch difference between the front harmony and the one after it (pitch progression from the front harmony) ing. Besides this, information indicating weak or strong beats,
Information that distinguishes note lengths may be added to the set of functions F.
Flow chart of calculation of each function F (Figs. 16 to 23)
(Fig.) Is clear from the description of itself, and therefore its explanation is omitted.

FIG. 13 shows the details of forward reasoning in 33-7 in FIG. First, the rule number pointer P is set to "1" indicating the rule that is the root of the production rules (44-1). Then, it is checked whether or not the premise part (Lp ≦ Fxp ≦ Up) of the rule indicated by the rule number pointer P is satisfied. If not, the data Np of the negative conclusion part of the rule is used as a pointer to the next rule.
However, when the data Yp and Np are negative values, the final conclusion has been reached, so the absolute value (−Yp, −Np) is used as the conclusion register as the non-harmonic type identifier or the illegal sound identifier. Set to. According to the flow, the condition Lp of 44-3>
When the condition Fxp> Up of Fxp or 44-5 is satisfied, the premise part Lp ≦ Fxp ≦ Up of the rule P is not satisfied, so the data Np of the negative conclusion part of the rule is substituted into a (44-4, 44 -6),
In other cases, since the premise part is established, the data Yp of the affirmative conclusion part of the rule P is entered in a (44-2). Substituting this a into P (44-7), when P is a positive value, it returns to the inspection of the next rule as the next rule number pointer, and when P is negative, the classification of non-harmonic sounds of -P The result (44-8, 44
-9).

Although the description of the embodiment is completed above, the present invention is not limited to the above embodiment, and various modifications and changes can be made.

For example, in the above-described embodiment, the chords stored in the chord memory 6 are exhaustively examined to determine whether or not the chords can be added to the melody, but the chord candidates may be input by the user. In this case, a learning machine for coding will be provided. At that time, when it is found from the classification knowledge of the non-harmonic sounds that the input chord is an unusable chord, the explanation may be displayed.
Japanese Patent Application No. 62-325178 (December 24, 1987) relating to the applicant
Japanese application) describes a technique for explaining the classification result of non-harmonic sounds (Fig. 71), and this technique can be used.

[Effects of the Invention] The automatic chord adding device according to the present invention, when a melody and a chord candidate are given, a knowledge storage means for storing music knowledge for classifying non-harmonic tones included in the melody according to a tone sequence of the melody. , An inference means that applies this musical knowledge to a given chord candidate that is assumed to be a melody to classify non-harmonic sounds included in the melody, and a chord candidate that is assumed according to the result of the inference means is converted into a melody. There is provided a judgment means for judging whether or not there is a usable code. Therefore, there is an advantage that all chord candidates that can be explained musically are determined to be usable. Also, the length of the given melody is not limited,
It can also be used for melodies that include modulation.

[Brief description of drawings]

FIG. 1 is an overall configuration diagram of an automatic chord adding device according to an embodiment of the present invention, FIG. 2 is a general flow showing the operation of the embodiment, and FIG. 3 is a diagram showing an example of a melody and chord data format, FIG. 4 is a diagram showing an example of melody data, FIG. 5 is a flowchart for reading melody data, FIG. 6 is a flowchart for obtaining the first (Ps) and last (Pe) of the melody data in the section to be inspected, and FIG. 7 is 8 is a flowchart for calculating Ps, FIG. 8 is a flowchart for calculating Pe, and FIG. 9 is a flowchart for reading production rules.
Figure 10 shows an example of knowledge of production rules.
Figure 11 is a flow chart for reading code data, No. 12
Figure is a flowchart for judging whether or not chord candidates can be added. Figure 13 is a flowchart for classifying nonharmonic sounds. Figure 14 is a flowchart for distinguishing harmony and nonharmonic sounds. Figure 15 is the situation of the melody to be analyzed. 16 is a flowchart for calculating the function F1, FIG. 17 is a flowchart for calculating the function F2, FIG. 18 is a flowchart for calculating the function F3, and FIG. 19 is a flowchart for calculating the function F4. FIG. 20, FIG. 20 is a flowchart for calculating the function F5, FIG. 21 is a flowchart for calculating the function F6, FIG. 22 is a flowchart for calculating the functions F7 and F8, and FIG. 23 is a temporary storage of the calculated function F. The flowchart, FIG. 24, is a flowchart for classifying nonharmonic sounds by forward inference. 1 ... CPU, 5 ... melody data memory, 6 ... chord memory, 8 ... production rule data memory,
RSi: an identifier indicating the result of classification of non-harmonic sounds, ill an identifier indicating an illegal sound.

Claims (1)

(57) [Claims] 1. An automatic chord adding device for automatically adding chords to a given melody, wherein a melody for inputting data representing a time series of notes constituting the melody (hereinafter referred to as a melody tone string). Input means, chord candidate giving means for giving chord candidates to a predetermined section of the melody, and knowledge memory for storing musical knowledge for classifying non-harmonic tones included in the melody in accordance with chords and melody tone strings. Means for applying the musical knowledge to the melody and chord candidates in the predetermined section, that is, assuming that the chord candidate is a chord for the melody in the predetermined section, and is not a constituent sound of the chord candidate. Inference means for performing classification of non-harmonic sounds on melody sounds, and when the classification result of the inference means is a predetermined classification result, the chord candidates are set to the predetermined section. It determines that the chord with respect to the melody, the automatic chord additional apparatus characterized by having a the determining means not chord for melody of the predetermined interval candidate 該和 sound when not predetermined classification results.