JPH0981145A - Musica performance data analyzer and musical performance data analysis method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To detect key from the melody of performance data. SOLUTION: Melody data is divided into phrases (#1 to #N) and the respective phrases are divided into chord rhythms (#1 to #1M, to, #N1 to #NL) of one measure or two beats according to the characteristics of the melody. Chord tones are extracted from the respective chord rhythm sections. The chord rhythms are combined at need in accordance with the extracted chord tones. The chord rhythms (#11 to #1M', to, #N1 to #NL') changed in the sections are determined as the detection sections for detecting the key. The chord tones and pitch name strings are extracted by each of the respective detection sections. The keys which are candidates are extracted in accordance with the detected chord tones and pitch name strings. The key of the detection sections is determined from the extracted candidate keys.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus for analyzing performance data in electronic musical instruments, computer music and the like, and performance data analysis for detecting a key from melody information composed of data corresponding to notes of the performance data. The present invention relates to a device and a performance data analysis method.

[0002]

2. Description of the Related Art Conventionally, it is necessary to detect a key of performance data in order to add a chord (chord) or arrange a melody in an electronic musical instrument or computer music. As a method to detect the key of such performance data,
There is a method to detect the key by regarding the last note of the melody indicated by the performance data as the tonic, but this method is not applicable when transposing in the middle of the melody, and it does not change the key in detail enough to determine chord progression. There is a problem that it cannot be detected. Therefore, there is a method of dividing the melody into blocks each having a predetermined length such as a phrase and detecting a key in each block, but it is still insufficient and accurate key detection cannot be performed.

[0003]

SUMMARY OF THE INVENTION An object of the present invention is to perform accurate key detection from melody information of performance data.

[0004]

In order to solve the above-mentioned problems, a performance data analysis apparatus according to claim 1 of the present invention comprises a performance data input means for inputting performance data indicating a sequence of notes, and a plurality of performance data input means. In each key, at least one of the note name string storage means for storing the note name string consisting of the main note names characterizing each key and the plurality of note names included in the performance data input by the performance data input means. A note name string including a part, a note name string extracting means for extracting from the note name string storing means, and a key detecting means for detecting a key based on the note name string extracted by the note name string extracting means; It is characterized by including.

According to a second aspect of the present invention, there is provided a performance data analyzing device, comprising: performance data input means for inputting performance data indicating a sequence of notes; and a main note name characterizing each key in a plurality of keys. And a note name sequence storing means for storing a note name sequence that stores at least a part of a plurality of note names included in the performance data input by the performance data inputting means. Means for extracting a note name string extracted from the means, a chord sound extracting means for extracting a harmony sound included in the performance data input by the performance data input means, and a note name string extracted by the note name string extracting means. Key detecting means for detecting a key on the basis of the key detecting means.

According to a third aspect of the present invention, there is provided a performance data analysis device for inputting performance data indicating a note sequence, and performance data input by the performance data input means in a predetermined section. Section dividing means for dividing into a plurality of tones, a note name string storage means for storing a note name string consisting of main note names characterizing each key in a plurality of tones,
A note name string extracting means for extracting from the note name string storing means a note name string containing at least a part of a plurality of note names included in the performance data of the section divided by the section dividing means; Key detecting means for detecting a key for each section based on the note name string extracted by the name string extracting means.

According to a fourth aspect of the present invention, there is provided a performance data analysis device for performing performance data input means for inputting performance data indicating a note sequence, and embroidery data from the performance data input by the performance data input means. An embroidery sound detecting means for detecting a sound pattern and a part of the detected embroidery sound patterns are determined to be a harmony sound, and the rest of the embroidery sound patterns are changed to a non-harmonic sound as an embroidery sound. Harmonic sound determining means for determining and determining performance data having a predetermined pitch relationship with the determined harmonic sound as a harmonic sound,
It is characterized by having.

In the performance data analyzing method according to the fifth aspect of the present invention, in a plurality of tones, a note name string consisting of main note names that characterize each key is stored in the note name string storing means, Performance data indicating a sequence of notes is input, and a note name sequence in which at least a part of a plurality of note names included in the input performance data is extracted from the note name sequence storage means, and is extracted. The key is to detect the key based on the note name sequence.

Further, in the performance data analyzing method according to the sixth aspect of the present invention, in a plurality of tones, a note name string consisting of main note names that characterize each key is stored in the note name string storing means, Performance data indicating a sequence of notes is input, a note name string in which at least a part of a plurality of note names included in the input performance data is extracted from the note name string storage means, and is input. The harmony sound included in the performance data is extracted, and the key is detected based on the extracted note name sequence and the harmony sound.

In the performance data analyzing method according to the seventh aspect of the present invention, a note name sequence storing means stores a note name sequence consisting of main note names that characterize each tone. Performance data indicating a sequence of notes is input, the input performance data is divided into predetermined sections, and a plurality of note names included in the performance data of the divided sections are at least partially included in a note name string. Is extracted from the note name string storage means, and the key for each section is detected based on the extracted note name string.

Further, in the performance data analysis method according to the present invention, performance data indicating a sequence of notes is input, an embroidery sound pattern is detected from the input performance data, and the detection is performed. A part of the embroidery sound pattern is determined to be a harmony sound, and the remaining part of the embroidery sound pattern is determined to be a non-harmonic sound as an embroidery sound, and a predetermined pitch relationship is established with the determined harmony sound. It is characterized in that certain performance data is determined as a harmony sound.

The key in the performance data is detected based on the harmony tone and the note name sequence in the performance data. Since the chord sound and the note name sequence represent the characteristics of the key of the melody, the chord is detected from the note name sequence extracted by the note name sequence extraction means, and the chord sound extracted by the harmony tone name sequence extraction means is detected. If the key is detected from the note name sequence, the accurate key can be detected.

[0013]

DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 2 is a block diagram of an electronic musical instrument to which the performance data analysis device of the embodiment of the present invention is applied.
U1 controls the entire electronic musical instrument by using the working area of the working memory 3 based on the control program stored in the program memory 2, operates the operator 4 such as a panel switch, and displays on the display device 5. Due to
Operates in normal mode or melody analysis mode.

In the normal mode, the CPU 1 fetches a key code, touch data, a key-on signal (or a key-off signal) corresponding to a key event of the keyboard 6 and outputs the key code and note-on or note-off to the sound source 7. Performs sound generation and mute processing corresponding to keyboard performance. The sound source 7 generates a tone signal corresponding to the data input from the CPU 1, and the sound system 8 performs D / A conversion and amplification of the tone signal to generate a tone.

In the melody analysis mode, first, the CPU 1
Takes in a key code, touch data, a key-on signal (or a key-off signal), etc. corresponding to a key event on the keyboard 6, and generates performance data based on these data.
It should be noted that the processing for generating the performance data is performed by inputting each note by the mode of the real-time record for generating the performance data in response to a normal performance operation by the keyboard 6 or by operating the keyboard 6 and the operator 4. Perform in step record mode.

Further, from this performance data, the variation of the sound length at the time of the input performance is corrected (quantized) to generate the sound length data indicating the note length of the note, and the pitch data, the sound length data,
Melody data composed of rest data and rest length data is constructed and stored in the data memory 9 (RAM). And
When the melody data is stored, the CPU 1 analyzes the stored melody data as described later in detail, and divides the melody data into segment analysis data, detection tone data, and detection chord (chord) data. Is stored in the data memory 9.

FIG. 3 is a diagram showing each format of the melody data and the section analysis data in the analysis data. Same figure
In the melody data shown in (A), one note is represented by pitch data corresponding to the pitch and note length data corresponding to the note length of the note, and rest data and rest length corresponding to the note length of the rest. The data represents one rest. The pitch data is a key code, and the rest data is a predetermined specific identification code. The note length data and the rest length data are data in which the time length is represented by the number of clocks with a predetermined clock (for example, a clock obtained by dividing a quarter note into 24 pieces) as a unit.

In the actual performance, the actual sounding time may be different between playing the same note with slur and staccato, but as described above, the above note length data and rest length data are used. Is processed so that the number of clocks corresponds to the note length regardless of the actual pronunciation time. Therefore, by adding the note length data and the rest length data from the beginning of the melody data, the position of the bar line is identified, and thus the melody data can be searched in bar units or phrase units.

The section analysis data shown in FIG. 3B is obtained by dividing the melody data into chord rhythm sections according to the melody analysis result, as will be described later.
It is recorded in the format in which the divided data indicating the boundary of each chord rhythm section is inserted in the melody data of (A). The end data is recorded at the end of the melody data and the section analysis data.

FIG. 1 is a conceptual diagram for explaining the flow of melody analysis in the embodiment. First, the melody data of the data memory 9 is divided into phrases (# 1 to #N), and each phrase is divided into more detailed chord rhythms (# 11 to # 1M, ..., #N).
1 to #NL). Here, the phrase is a unit section when the melody data is divided into predetermined lengths with good delimitation, and a method of this division will be described later. A chord rhythm is a section smaller than a phrase, such as one bar or two beats, and means a unit section occupied by one chord in chord progression. The way of this division is determined according to the characteristics of the melody as described later.

Next, a chord rhythm is extracted from each chord rhythm section, chord rhythms are combined as needed to form a chord detection detection section based on the chord rhythm, and chords whose sections are changed. Rhythm (# 11 ~ # 1M ', ~, # N1 ~ #N
L '). A harmony tone and a note name sequence are extracted for each detected section of the reconstructed chord rhythm, and a candidate key is extracted based on the harmony tone and the note name sequence.

In this embodiment, for example, as shown in Table 1 below, note name strings consisting of major note names that characterize the key corresponding to each key scale are represented as I, II, III, and IV. It is stored in the program memory 2 as a scale tone sequence table with various frequencies, and when a candidate key is extracted from the harmony tone and the tone name sequence, the scales included in the tone name sequence of the detected section from this scale sequence table. A tone string is searched for, and a key of a tone scale including this tone string and the harmony sound of the detected section is set as a candidate tone. Then, the key of the detected section is determined from this candidate key.

[0023]

[Table 1]

In this embodiment, the chord (chord) of the detection section divided from the detection section divided for key detection and its chord sound
The chord is detected based on the chord progression data stored in the chord sequence DB (database) 10 (FIG. 2). At this time, the chord is decided by matching with the chord progression data while inspecting whether the decided chord is suitable as the beginning or end chord of the phrase.

FIG. 5 is a diagram showing an example of chord progression data. The chord progression data indicates a typical chord progression, and is composed of 4 chord progressions, 3 chord progressions, and 3 chord progressions.
There are three chord progressions consisting of one chord and two chord progressions consisting of two chords. The 4 chord progression, the 3 chord progression, and the 2 chord progression are classified into chord progression data corresponding to the major and the minor, respectively.
Further, the four chord progressions are classified into a synchronous type in which one chord corresponds to one bar and an asynchronous type in which no chord corresponds to one bar. Each chord of these chord progressions is indicated by a combination of root note (I) and type (maj) like Imaj. Here, since the root note is represented by a frequency based on the tonic of the key, the chord progression data is 1
The tones of two tones (C, C #, D, to, B) can be commonly handled.

Further, the chord progression data of the same kind are given priority levels. For example, in the case of the major of the 4 chord progression (synchronous) DB shown in FIG. 5, “Imaj → VImin → II
m7 → V7 ”(in C major, Cmaj → Amin → Dm7 → G7
) Has the highest priority level, and “IIIm7 → IVmin → IIm7 → V
7 "(Em7 → Fmin → Dm7 → G7 in C major)," I
maj → IIm7 → V7 → Imaj ”(in C major, Cmaj → Dm7
→ G7 → Cmaj), and the priority levels are lower.

FIG. 6 is a flowchart of the melody analysis process of the control program in the melody analysis mode in the embodiment, and FIGS. 7 to 17 are flowcharts of the subroutine. The operation of the embodiment will be described based on each flowchart.

When the melody analysis mode is designated by the operation of the operator 4, the melody analysis process of FIG. 6 is started,
In step S1, the data input process of FIG. 7 is performed to store the melody data, and if there is a chord rhythm input, it is also stored. If not, the chord rhythm automatic setting mode is set. In step S1-1, it is determined whether or not the chord rhythm automatic setting mode is set, and if it is the chord rhythm automatic setting mode, step S2
Otherwise, go to step S4. Step S2
Then, the section temporary division processing of FIG. 8 is performed to divide the melody data into phrases. Next, in step S3, the chord rhythm section extraction process of FIG. 9 is performed to divide each phrase into chord rhythm sections. In step S4, the chord sound extraction process of FIG. 10 is performed, and in step S5, the chord sound re-extraction of FIG. 13 is performed. A chord rhythm is extracted by performing processing, and chord rhythm combination processing of FIG. 14 is performed in step S6 to determine a chord rhythm section as a detection section for key detection. Then, step S7
Then, the key detection processing of FIG. 15 is performed to fix the key. Further, in step S8, the code is determined by performing the code detection process of FIG.

In the data input process of FIG. 7, step S1
At 1, the melody is input based on the performance operation from the keyboard 6, and the melody data is stored in the data memory 9. Next, in step S12, manual input processing of time signature data such as 4/4 time signature and 3/4 time signature is performed by the operator 4, and there is no manual designation input of the chord rhythm section by the operator 4 in step S13. In this case, the chord rhythm is automatically set in the automatic setting mode. If there is a manual designation input, the delimiter of the designated chord rhythm section is stored.

In the section temporary division processing of FIG. 8, step S2
In 1, when a note equal to or longer than a half note (a note having a note length longer than the note length of a half note) is immediately before the bar line, the position of the bar line is set as the phrase division position. When the phrase has four or more measures, the melody data is re-divided into phrases with the position of the bar line every four measures from the front as the phrase dividing position. At this time, as in the case of FIG. 3B, the division position is stored by inserting the division data of the phrase different from the chord rhythm into the melody data.

Here, in the chord rhythm extraction process of FIG. 9 which will be described later, each of the stored phrases is further divided into chord rhythm intervals of one bar or two beats according to the melody pattern. The inspection condition of is switched according to the note density in the phrase. Therefore, the flag DV corresponding to each phrase
(FN) (FN is the phrase number) is used to detect the flag DV (F
N) is set to "1" and the note density is high, the flag DV (F
N) is stored as “2”.

First, in step S22, it is determined whether or not the currently set time signature is three quarters, and three quarters is determined.
If it is a time signature, all flags DV (all) corresponding to all the phrases are set to "1" in step S23 and the routine returns to the original routine. If it is not three quarters, the read pointer ( The read pointer in the data memory 9. The same applies hereinafter) is set and the first phrase is noted and the process proceeds to step S25.

In step S25, it is determined whether or not the ratio of the total note length of the eighth note or more among the notes in the phrase to the phrase length is 0.6 or more, and if it is 0.6 or more. For example, the flag DV (FN) is set to "1", and otherwise the flag DV (FN) is set to "2". Then, in step S26, it is determined whether or not it is the last phrase at present, and if it is not the last phrase, the read pointer is updated to the next phrase in step S27, and step S2 is performed.
If it is the last phrase, the process returns to the original routine.

By the above processing, when the ratio of the total note length of the eighth note or more and the phrase length is 0.6 or more in the case of the three-quarter time signature and other than the three-quarter time signature, the phrase If there are many notes with long note lengths (low note density), and the ratio of the total note length of 8th notes or more to the phrase length is less than 0.6, except for the third quarter time signature, It is considered that there are many notes with a short note length (the note density is high), and in the chord rhythm extraction process of FIG. 9, the inspection condition is switched according to the note density to inspect the melody pattern.

In FIG. 9, first, in step S31, the read pointer is set to focus on the first measure, and in step S32 it is determined whether DV (FN) = 1.
If DV (FN) = 1 (note density is low), it is determined in step S33 whether the notes included in the current measure are two notes or less. If two notes or less, step S37.
Insert the divided data with and divide into chord rhythm sections of one measure. If it is not less than two notes in step S33, the trill pattern is searched in step S34 to check whether the trill pattern covers the entire one bar. Note that the trill pattern is a pattern in which the musical notes of the pitch a and the pitch b (the pitch difference between a and b is within 1 note) repeats abab ...

If the trill pattern of one bar is determined in step S35, the chord rhythm section of one bar is divided in step S37. If it is not one trill pattern, the process proceeds to step S36. In step S36, it is determined whether or not all the notes of one measure are quarter notes, and if all the notes are quarter notes, it is divided into chord rhythm intervals of one measure in step S37, and otherwise. In step S305, the divided data is inserted into the bar to divide into chord rhythm sections of two beats, and the process proceeds to step S306.

On the other hand, DV (FN) = 1 in step S32.
If not (note density is high), in step S38, as shown in FIG.
The special pattern shown in is searched in the current bar, and it is determined in step S39 whether or not this special pattern exists. This special pattern is a sounding timing pattern of two beat lengths appearing from the first beat to the second beat or from the third beat to the fourth beat, and a pattern extending over two beats and three beats is ignored. This special pattern is 1 for 2 beats that are frequently used
It is organized. In addition, both the original melody pattern of the melody data which is initially input and the case of the sounding timing pattern which is generated as a result of collecting continuous tones having the same pitch as one tone are checked. Figure 4 (A)
In the case of, the condition is that the processes proceed sequentially and only in the same direction. It should be noted that “sequential progression” means that the pitch difference between adjacent notes progresses up or down by 2 degrees (whole tone or semitone), and the same direction means one direction in which the pitch rises or falls. . Further, in the cases of FIGS. 4 (B) and 4 (c), there is no condition.

If there is this special pattern, it is a pattern of 2 beats in length, so it is divided into chord rhythm sections of 2 beats each in step S305. Search for 4-trill trill patterns. Then, step S30
If there is a trill pattern of 4 beats in the judgment of 2, then step S
At 37, it is divided into chord rhythm sections of one measure, and if there is no trill pattern of 4 beats, it is determined at step S303 whether or not they all have the same phonetic name. If all have the same note name, it is divided into chord rhythm sections of one measure in step S37, and if at least one is a different note name, it is determined in step S304 whether or not there is a note with a note length of eighth note or less. And 8
If there are no notes below the quarter note, the chord rhythm section of one bar is divided in step S37, and if there are notes below the eighth note, it is divided into chord rhythm sections of two beats in step S305.

When the processing for one bar is completed, it is determined in step S306 whether the currently focused bar is the last bar. If it is not the last bar, the read pointer is moved to the next bar in step S307. After updating to the measure, the process proceeds to step S32, and if it is the last measure, the process returns to the original routine.

By the above processing, if the trill pattern extends over one bar, if all the notes in one bar are quarter notes, if there is a trill pattern of four beats in one bar, then all the same names in one bar. If there is no key change (or chord change) within one measure, such as when the measure is broken, the chord rhythm section is divided into one measure. A chord rhythm with a 2-beat length is used when there is a possibility that there will be a change in key within one measure, such as when there is a special pattern that is a 2-beat length, there is no long trill pattern, or when there are scattered note names. It is divided into sections. In this way, each phrase is provisionally divided into chord rhythm intervals of one measure or two beats according to the melody pattern and the note length. In addition, when the note of the original melody pattern spans two chord rhythm sections when dividing into the above chord rhythm section, the note length of the note (the last note of the previous chord rhythm section) is cut at the dividing position. Then
Tie information (see FIG. 3 (B)) indicating the remaining chord length is added before the first note in the subsequent chord rhythm section.

Next, the chord sound in the chord rhythm section is extracted from the beginning of the melody by the chord sound extraction process of FIG. 10 (and the chord sound extraction rule process of FIG. 11 and the pre-taken sound process of FIG. 12), and the chord sound is extracted. With respect to the chord rhythm section which could not be performed, the harmony tones are re-extracted in order from the back of the melody by the harmony tones re-extraction processing of FIG. Therefore, a detection flag indicating whether or not a harmony sound is detected corresponding to each chord rhythm section is used, the undetected state of the harmony sound is stored as a detection flag “0”, and the detected state of the harmony sound is detected. It is stored as "1".

In the chord sound extraction process of FIG. 10, step S
At 41, the read pointer is set to pay attention to the first chord rhythm section, at step S42, the detection flag of the chord rhythm section of interest is set to "0" (not detected) in advance, and the routine proceeds to step S43. In step S43, when the beginning of the chord rhythm section of interest is a rest, there is a note in the immediately preceding chord rhythm section, and the note is not a harmony,
This rest is absorbed by the preceding note. That is, the rest data is deleted and the rest note length is added to the note length of the immediately preceding note to obtain the note length data.

Next, in step S44, a pattern including a melody, that is, an embroidery sound, in which the notes of the pitch a and the pitch b are arranged in aba in the focused chord rhythm section is detected (between ab and ba). Even if there is a rest, the rest is ignored.) It is determined in step S45 whether or not the embroidery sound pattern is detected. If the embroidery sound pattern is detected, the note of pitch a is set as a chord sound and the note of pitch b is set as an embroidery sound in step S46, and the pitch a is set once in the note of the chord rhythm section of interest in step S47. When there is a third tone, there is no sixth tone, and when there is no additional fourth tone, the priority level is searched for, and the sound detected at this priority level is set as a harmony tone. Then, in step S48, other notes are set to other non-harmonic sounds, in step S49, the detection flag of the chord rhythm section of interest is set to "1", and in step S402.
Proceed to.

As described above, while focusing on the embroidery sound pattern to detect the harmony sound, if the embroidery sound pattern is not detected at step S45, the harmony sound extraction rule process of FIG. 11 is performed at step S401 to perform the embroidery sound pattern. Harmonic tones are extracted by rules other than, and the process proceeds to step S402. Then, when the processing for one chord rhythm section is completed, it is determined in step S402 whether or not the current chord rhythm section of interest is the last chord rhythm section. If it is not the last chord rhythm section, it is read in step S403. Update the pointer to the next chord rhythm section and step S
In step 42, if it is the last chord rhythm section, it returns to the original routine.

In the chord sound extraction rule process of FIG. 11, it is determined in step S410 whether or not the chord rhythm section of interest includes only rests, and if only rests, the process returns to the original routine and rests only. If not, it is determined in step S411 whether or not there is one kind of note name. If there is only one kind of note name, all notes are harmonized in step S412, and the flow advances to step S418.

On the other hand, if there is not one kind of note name in step S411, the pre-taken sound processing of FIG. 12 is performed in step S413. Here, the pre-taken sound process of FIG. 12 is a process for a note called anticipation in which a constituent sound of a subsequent chord precedes a final beat of the preceding chord when the chord changes, for example. First, in step S420, it is determined whether or not the condition "the beginning of the next chord rhythm section is a note and the note length is a quarter note or longer" is satisfied. If the condition is not satisfied, the original routine is directly returned to. If the condition is satisfied and the condition is satisfied, in step S421, if the last note of the current chord rhythm section of interest is a note, the note is returned to the original routine as the first note of the next chord rhythm section. As a result, the anticipation notes are arranged in the same chord rhythm section.

When this pre-acquired sound processing is completed, it is determined in step S414 in FIG. 11 whether or not the number of notes in the chord rhythm section of interest is three notes.
At 415, the two tones having the largest pitch difference among the three tones are set as the harmony tones, and the process proceeds to step S418. If it is not three notes, it is determined in step S416 whether or not there is a note having a note length of ½ or more of the chord rhythm section length (hereinafter referred to as note i). If there is no note i, the process returns to the original routine as it is. If there is a note i, the note i is made into a chord sound in step S417, and if there are other chord forming notes of chords having the note i as chord forming notes, those notes As a chord sound and the process proceeds to step S418. Then, in step S418, the other notes are set to other non-harmonic sounds, and in step S419, the detection flag of the chord rhythm section of interest is set to "1" and the process returns to the original routine.

As described above, the chord sound extraction process of FIG. 10, the chord sound extraction rule process of FIG. 11, and the pre-taken sound process of FIG. 12 extract the chord sound in the chord rhythm section from the beginning of the melody, and the chord sound is extracted. The detected flag is "1" for the chord rhythm section that has been played, and the detection flag is "0" for the chord rhythm section in which the chord sound is not extracted. Therefore,
The chord sound re-extraction processing of FIG. 13 is performed from the end of the melody to re-extract the chord sound for the chord rhythm section in which the detection flag is "0".

First, in step S51, the read pointer is set to focus on the last chord rhythm section, and in step S52 it is determined whether or not the detection flag = 0 of the chord rhythm section of interest. If the detection flag is not 0, a harmony sound has already been detected, so the flow proceeds to step S501.
If the detection flag = 0, no harmony sound has been detected, so a harmony sound is detected in step S54 and subsequent steps.

In step S54, the condition that the chord rhythm section of interest is "two notes, the beginning note of the following chord rhythm section is a chord note, and these three notes are progressive in the same direction" is satisfied. If it is determined that the condition is satisfied, in step S55, the first sound of the three sounds is set as a harmony sound, and the process proceeds to step S59.

On the other hand, if the condition is not satisfied in step S54, the chord rhythm section of interest is "2" in step S56.
It is a sound and the sound in the subsequent chord rhythm section is one sound "is determined, and if the condition is satisfied, the second note is set as a harmony sound in step S57 and the process proceeds to step S59. If the condition is not satisfied, in step S58, the pitch difference between any two notes in the chord rhythm section of interest is 3 degrees, 6 degrees if none, or 4 if not.
Whether or not there is a degree is checked at the priority level, and two tones having the pitch difference detected at the priority level are set as a harmony sound. If there are two or more sets having the same frequency of the highest priority level among the sets having these pitch differences, the bass side is prioritized. Then, step S5
At 9, the other notes are made into other non-harmonic sounds.

When the detection of the harmony sound for one chord rhythm section is completed as described above, it is determined in step S501 whether or not the chord rhythm section currently focused on is the first chord rhythm section. If it is not the chord rhythm section, the read pointer is updated to the immediately preceding chord rhythm section in step S53 and the process proceeds to step S52. If it is the first chord rhythm section, the original routine is directly returned.

By the above processing, the chord rhythm interval of the chord rhythm section obtained by temporarily dividing the phrase into one bar or two beats is extracted. Next, since this temporary division into chord rhythm sections may be divided more than necessary for key detection, the chord rhythm combination processing of FIG. The process of combining is performed based on the harmony sound.

In the chord rhythm combining process of FIG. 14, the read pointer is set in step S61 to focus on the first bar, and in step S62 it is determined whether or not the bar of interest is divided into chord rhythm intervals of two beats. To do. If it is not divided into chord rhythm intervals of two beats, step S
At 64, one bar is set as a detection section for the bar of interest, and the process proceeds to step S66.

On the other hand, if the bar of interest is divided into chord rhythm intervals of two beats, in step S63, "the chords of the constituent sounds that match all the chords of the bar of interest are present and are set. It is determined whether or not the condition that the tempo is quarter note> 46 (tempo exceeding 46 beats per minute when one quarter note is one beat) is satisfied. If the condition is satisfied, one bar is set as the detection section for the target bar in step S64, and if the condition is not satisfied, every two beats are set as the detection section for the target bar in step S65. When the processing for one bar is completed, it is determined in step S66 whether or not the target bar is the last bar, and if it is not the last bar, step S66.
In 67, the read pointer is updated to the next bar, and step S
The process proceeds to 62, and if it is the last measure, the original routine is directly returned to.

In the chord rhythm combining process described above, when the bar is provisionally divided into chord rhythm intervals of two beats, there are chords of constituent sounds that match all the harmony sounds of the bar. In addition, if the tempo is relatively fast, it is considered that there is a high possibility that the key will not change within that bar, and the chord rhythm section for every two beats is grouped into one bar to be the detection section. There is.

As described above, when the melody data is divided into one bar or two beats and the detection section is determined, the key detection is performed by the key detection processing of FIG. First, in step S71, the read pointer is set to pay attention to the first detection section, and in step S72, it is determined whether the key of the immediately preceding detection section is fixed. If the key of the immediately preceding detection section is not fixed (including the case where the target detection section is the first detection section and the immediately preceding detection section does not exist), the candidate key extraction processing of FIG. The candidate tone of the section is extracted, and the process proceeds to step S77.

On the other hand, if the key of the immediately preceding detected section has been confirmed in step S72, it is determined in step S73 that there is an out-of-key tone (sound that is not in the scale of the key) for the determined key in the noticed detected section. If there is an out-of-control sound, step S
76, if there is no out-of-control tone, the key of the noticed detection section is decided as the key of the immediately preceding detection section in step S74, and the step S74 is executed.
Proceed to 705.

Here, the candidate tone extraction processing of FIG. 16 will be described first. First, in step S81, the note names of the notes in the target detection section are picked up and sorted, and in step S82,
From the scale sequence table of Table 1 above, a scale sequence containing three or more note names in the sorted sequence of tone names is searched, and step S8 is performed.
In step 3, if there is a matched scale string, a scale key that includes each matched scale string and the harmony sound of the target detection section is searched, and the detected key (multiple keys if there are multiple keys) is detected. It is stored as a section candidate tone. Then, step S84
Then, from the stored candidate tones, the minors having the fifth and seventh minors of the key in the noticed detection section are excluded from the candidate tones, and the process returns to the original routine. When the candidate tones are extracted as described above, the tones are determined if the predetermined conditions are met after step S77 of FIG.

First, in step S77, the condition (A) "candidate key of major (maj) that has 1st and 7th tones of the scale of the key from the detection section where the key is not fixed to the target detection section" Then, in step S78, the condition (B) “a minor (minor) with 1st, 5th and 7th notes of the scale of the key from the detection section where the key is not fixed to the target detection section Pick a candidate tone, and in step S79, condition (C)
"There are 1st note, 6th note and 7th note in the scale of the key from the detection section where the key is not fixed to the target detection section (min
or)).

Next, in step S701, conditions (A) to (C)
It is determined whether or not there is a candidate tone picked up in step S702, and if there is no candidate tone, it is determined in step S702 whether or not the target detection section is the last detection section. If it is the last detected section, in step S703, the section from the undetected detected section to the noticed detected section is fixed to the key determined immediately before and the process returns to the original routine. Without confirming the key, the process proceeds to step S75.

On the other hand, if there are candidate tones picked up under the conditions (A) to (C), in step S704, the candidate tones are most frequently cited from the detection section where the key is not fixed to the noticed detection section. The key is determined and stored as the key from the unfixed detection section to the target detection section, and the process proceeds to step S705. When there are a plurality of keys mentioned most, the same key as the key of the preceding detection section (finally fixed detection section) and the relative priority are determined. If the previous key has not been confirmed, (A) (B) (C)
Set to the key picked up in the priority level of.

When the process for one detection section is completed, it is determined in step S705 whether or not the target detection section is the last detection section. If it is not the last detection section, the read pointer is moved to the next step in S75. The detection section is updated to step S72, the process proceeds to step S72, and if it is the last detection section, the process returns to the original routine.

As described above, the melody is divided into detection sections of two beats or one measure according to the harmony sound and the melody pattern, and the candidate tone is extracted for each detection section by the harmony sound and the note name sequence. The key is determined for each detection section based on the key. In addition, in steps S77 to S79, step S703, and step S704 of FIG. 15, "from the detection section in which the key is not fixed to the target detection section" is the target detection when the key is set up to the immediately preceding detection section. When only the section is referred to and at least the key can be determined, the key is determined for each detected section.

In the code detection process of FIG. 17, step S
At 91, processing is performed when there is a portion where four or more detection sections are continuous in units of one bar (including the case where all are in units of one bar). That is, the four codes of the four chord progressions (synchronization) in the chord progression data are associated with the four detection sections,
If the harmony sound of each detection section is included in the constituent sounds of the corresponding chord, and if one of the detection sections is the chord suitable for the beginning or end of the phrase if it is the beginning or end of the phrase, then the four detections The chords of the section are determined as four chords of the four chord progressions (synchronization).
At this time, which of the DB for the major key and the DB for the minor key is used is determined by the key (major or minor) obtained as described above, and is expressed by the frequency by considering the tonic of the key. Chords can be converted into real chords. When there are a plurality of such 4 chord progressions (synchronizations), the decision is made according to the priority level. A code group suitable for the beginning or the end of the phrase is stored in the program memory 2 in advance, and it is determined whether or not it is suitable by comparing it with them. Here, if it is assumed that the four codes in the four chord progressions (synchronization) are associated with the four detection sections in the four detection sections, the next four new detection sections
The same process is performed for each detection section. On the other hand, when the correspondence between the four codes and the four detection intervals cannot be obtained,
One of the four detection sections is moved backward, and the same processing is performed for a total of four sections including the last three sections of the previous time and a new one section. By repeatedly executing such processing, the case where the code of 4 chord progression (synchronization) is applied to all the detection sections is determined.

Next, in step S92, processing is performed on the detection section for which the code could not be determined in step S91. That is, four chord progressions in the chord progression data (asynchronous) with respect to the detection section in which the chord is not determined (the detection section which is not the processing target in step S91).
4 chords corresponding to the four detection intervals are included, and the harmony sounds of the respective detection intervals are included in the constituent sounds of the corresponding chords, and if one of the detection intervals is the beginning or end of the phrase, When it is the last suitable code, the codes of these four detection intervals are determined as the four codes of the four code progressions (asynchronous). If there are a plurality of such 4 chord progressions (asynchronous), it is determined according to the priority level. Here, asynchronous means that four chord progressions do not necessarily correspond to four detection intervals. That is, one code may extend over a plurality of detection intervals. For example, assume that there is an asynchronous code progression of C → Am → Dm7 → G7. This chord progression may be assigned to eight detection intervals and assigned as C → C → C → C → Am → Am → Dm7 → G7. Therefore, the above-mentioned step S9
When the asynchronous chord progression is made to correspond to four or more consecutive detection sections that could not be determined in step 1, it is first judged whether or not there is a chord progression corresponding to the section in the first four sections, and if there is, the code If it is decided to proceed, if there is not, the section is increased by 1 to make 5 sections. Then, the five sections are arbitrarily grouped into four sections, and the grouped four sections are judged to match the chord progression. Thereafter, each time it is determined that there is no matching chord progression, the section is increased by one and the same processing is performed. If no matching chord progression is found even after reaching the end of the continuous section, the section is shifted backward by one and the same processing is performed.
This is repeated until the last four sections of the continuous section are reached. If no matching chord progression is found, the process proceeds to the next step S93.

Next, in step S93, a process is performed for a part where at least three detection intervals are continuous in the part where the code is not determined. That is, three chords of three chord progressions in chord progression data and three detection zones are associated with a detection zone in which no chord is determined, and the chord sounds of the respective detection zones become the constituent sounds of the corresponding chords. When included, the chords of the three detection intervals are determined as the three chords of the three chord progressions, respectively. When there are a plurality of such 3 chord progressions, it is determined according to the priority level. In the case of 3 chord progressions, as in the case of 4 chord progressions (synchronization), if there is no correspondence with the detection section, the processing is repeated while shifting the detection section by one backward.

Next, in step S94, a process is performed on a part where at least two detection intervals are continuous in the part where the code is not determined. In other words, two chords of two chord progressions in the chord progression data are associated with two detection zones with respect to a detection zone in which no chord is determined, and the chord sound of each detection zone becomes a constituent sound of the corresponding chord. When included, the chords of the two detection intervals are determined as the two chords of the two chord progressions, respectively. In addition, when there are a plurality of such two-code progressions, it is determined according to the priority level. In the case of this 2 chord progression, the processing procedure is the same as in the case of 4 chord progression (synchronization).

Next, in step S95, 4 chord progressions, 3
The process is performed for the part where the chord is not decided in the chord progression and the 2-chord progression. That is, for each detection section for which a chord has not been determined, among the chords included in the chord tones of the chord corresponding to each detection section,
The code with the highest priority level is determined. By this processing, there is no code undecided section, and code decision of all sections is completed.

When determining the above chords, for example, chords competing with the melody as shown in Table 2 below are excluded. The table shows the case where the root note is "C", but the chords competing with the melody have the same pitch relationship as the table even when the root note is another root note.

[0071]

[Table 2]

In this embodiment, it is checked whether the chord is suitable for the beginning or the end of the phrase, but it may be checked whether the chord is suitable for the beginning or the end of the song. Further, the chord progression data may be stored as a database that varies depending on the rhythm type, tune style, tempo, time signature, and the like.

As described above, since the harmony sound and the note name sequence of the melody data are extracted and the key is detected from the scale containing the harmony tone and the note name sequence, the key detection faithful to the melody can be performed. it can. In addition, the melody data is divided into detection sections of two beats or one measure shorter than the phrase according to the harmony sound and the melody pattern, and a key is detected for at least the detectable detection section for each detection section. It can be performed.

In the embodiment, the chord rhythm section is set to 2
Although it is divided into beats or one bar, the length of this division may be switched according to the tempo.

Further, in the embodiment, the melody data is divided into phrases by the segment temporary division process, and is divided into the detection segments of one measure or two beats by the chord rhythm segment extraction process. It may be divided into (1 bar unit, 1 beat unit, etc.).

Further, in the embodiment, the performance data is input from the keyboard, but the performance data or the melody data may be supplied from the outside by MIDI data, for example. At this time, communication interfaces such as MIDI and various networks, FD and CD-RO
Melody data can be supplied from a storage medium such as M. Further, in the embodiment of the present application, the control program is stored in the program memory, but the program is stored in an external storage medium and the CP is stored if necessary.
It may be loaded into a device having U and used.

[0077]

As described above, according to the performance data analysis device and the performance data analysis method of the present invention, a note name sequence or a chord and a note name sequence are extracted from the input performance data, and the note name is extracted. Since the key in the performance data is detected based on the string or the harmony sound and the note name string, the accurate key can be detected. Also, since the performance data is divided into sections, key candidates corresponding to each section are extracted from the extracted note name sequence, and the key is determined from this key candidate, an accurate key can be detected.

[Brief description of drawings]

FIG. 1 is a conceptual diagram illustrating a flow of melody analysis according to an embodiment of the present invention.

FIG. 2 is a block diagram of an electronic musical instrument to which the performance data analysis device of the embodiment is applied.

FIG. 3 is a diagram showing each format of melody data and section analysis data in the example.

FIG. 4 is a diagram showing a two-beat length special pattern in an example.

FIG. 5 is a diagram showing an example of chord progression data according to an embodiment.

FIG. 6 is a flowchart of a melody analysis process in the embodiment.

FIG. 7 is a flowchart of a data input process in the embodiment.

FIG. 8 is a flowchart of a section temporary division process in the example.

FIG. 9 is a flowchart of chord rhythm section extraction processing in the embodiment.

FIG. 10 is a flowchart of a harmony sound extraction process in the embodiment.

FIG. 11 is a flowchart of a harmony sound extraction rule process in the embodiment.

FIG. 12 is a flowchart of a pre-taken sound process in the embodiment.

FIG. 13 is a flowchart of a harmony sound re-extraction process according to the embodiment.

FIG. 14 is a flowchart of chord rhythm combination processing in the embodiment.

FIG. 15 is a flowchart of a key detection process in the embodiment.

FIG. 16 is a flowchart of a code detection process in the embodiment.

FIG. 17 is a flowchart of a pitch conversion process for a bass pattern in the example.

[Explanation of symbols]

1 ... CPU, 2 ... Program memory, 3 ... Working memory, 4 ... Operator, 6 ... Keyboard, 9 ... Data memory, 10
... Code sequence DB.

Claims (8)

[Claims] 1. Performance data input means for inputting performance data indicating a note sequence, and note name sequence storage means for storing a note name sequence consisting of main note names characterizing each key in a plurality of keys. Note name string extracting means for extracting from the note name string storing means a note name string containing at least a part of a plurality of note names included in the playing data input by the playing data input means; A musical performance data analyzing apparatus, comprising: a key detecting unit that detects a key based on the note name string extracted by the column extracting unit. 2. Performance data input means for inputting performance data indicating a note sequence, and note name string storage means for storing a note name string consisting of main note names characterizing each key in a plurality of keys. Note name string extracting means for extracting from the note name string storing means a note name string containing at least a part of a plurality of note names included in the playing data input by the playing data input means; Harmonic sound extracting means for extracting a harmonic sound contained in the performance data input by the input means, and key detecting means for detecting a key based on the note name string extracted by the note name string extracting means. Performance data analysis device characterized by. 3. A performance data input means for inputting performance data indicating a sequence of notes, a section dividing means for dividing the performance data input by the performance data input means into predetermined sections, and a plurality of keys, At least a part is included in the note name string storage unit that stores a note name string consisting of main note names that characterize each key, and a plurality of note names included in the performance data of the section divided by the section dividing unit. A note name string extracting means for extracting a note name string from the note name string storing means, and a key detecting means for detecting a key for each section based on the note name string extracted by the note name string extracting means; A performance data analysis device comprising: 4. Performance data input means for inputting performance data indicating a note sequence, embroidery sound detection means for detecting an embroidery sound pattern from the performance data input by the performance data input means, and the detected. A part of the embroidery sound pattern is determined as a harmony sound, and the remaining part of the embroidery sound pattern is determined as a non-harmonic sound as an embroidery sound, and the determined harmony sound and a predetermined pitch relationship A performance data analysis device, comprising: a harmony sound determining means for determining the performance data in FIG. 5. In a plurality of tones, a note name sequence consisting of main note names that characterize each tone is stored in a note name sequence storage means, and performance data indicating a note sequence is input, A plurality of note names included in the performance data, a note name string at least a part of which is extracted from the note name string storage means, and a key is detected based on the extracted note name string. Characteristic performance data analysis method. 6. In a plurality of keys, a note name string consisting of main note names that characterize each key is stored in a note name string storage means, and performance data indicating a note sequence is input, A plurality of note names included in the performance data, a note name string including at least a part thereof is extracted from the note name string storage means, and a chord sound included in the input performance data is extracted; A performance data analysis device, characterized in that a key is detected on the basis of a reproduced note name sequence and a harmony sound. 7. In a plurality of keys, a note name string consisting of main note names that characterize each key is stored in a note name string storage means, and performance data indicating a note sequence is input, The performance data is divided into predetermined sections, and a note name sequence in which at least a part of a plurality of note names included in the performance data of the divided section is extracted from the note name sequence storage means, A performance data analysis method, wherein the key for each section is detected based on the generated note name sequence. 8. A performance data indicating a sequence of notes is input, an embroidery sound pattern is detected from the input performance data, and a part of the detected embroidery sound pattern is converted into a harmony sound. Along with the determination, the rest of the embroidery sound pattern is determined as a non-harmonic sound as an embroidery sound, and performance data having a predetermined pitch relationship with the determined harmony sound is determined as a harmony sound. How to analyze performance data.