JP3206370B2 - Music information analyzer - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a music information analyzer suitable for use in melody analysis and the like. ) And tones can be detected.

[0002]

2. Description of the Related Art In general, a tune may be composed of a single scale, a series of a plurality of scales (transposition / transposition), and a mixture of a plurality of scales. It may be composed of ambiguous scales. In order to cope with such a song, conventionally, an attempt has been made to detect the scale by calculating the appearance frequency of the pitch name for each section such as a bar.

[0003]

According to the above-mentioned prior art, scale detection is possible, but there are many problems in making the scale close to the scale actually judged by humans.

That is, since the appearance frequency of a pitch name is calculated for a short section such as a bar, there are many scales that are likely to be established, and it is difficult to determine the scale. Therefore, if the target section is lengthened, there is a possibility that the scale switching portion cannot be detected accurately.

There is a melody that makes the user feel a partially different scale for a moment. In order to cope with such a melody, it is necessary to take measures to avoid being distracted by partial modulation.

Further, if only the frequency of appearance of note names is used, a plurality of scales may be found at the same level, and measures for this need to be taken.

[0007] In short, in the automatic scale detection technology using a computer or the like, there are problems such as how to determine a target section, grasp of a scale flow (processing of partial modulation or ambiguous tone), processing of competing scales, and the like. .

Meanwhile, in a harmony minor scale or a melody minor scale, a detected scale component tone represents a minor tone. However, in the case of a natural scale, since there is a parallel tone, the tone is not determined only by detecting the scale component sound, and it is necessary to determine the key after determining whether the tone is a major or a minor.

It is an object of the present invention to provide a novel music information analyzing apparatus capable of detecting a scale or a key similar to a human sense.

[0010]

A first music information analyzing apparatus according to the present invention comprises: calculating means for calculating a degree of establishment of a plurality of scales for each phrase based on pitch information of a song divided into phrases; Extracting means for extracting a scale that satisfies a predetermined scale extracting condition for each phrase based on the degree of establishment calculated by the calculating means; and interposing between a plurality of phrases from which the same scale is extracted by the extracting means. And a search unit that searches for a phrase that satisfies a predetermined scale continuation condition from among the phrases for which the same scale as the plurality of phrases has not been extracted by the extraction unit, and sets the searched phrase to have the same scale as the plurality of phrases. Continuation means.

Further, a second music information analyzing apparatus according to the present invention comprises a calculating means for calculating the degree of establishment of a plurality of scales for each phrase based on the pitch information of a song divided into phrases, and the calculating means comprises: first extracting means for extracting a predetermined scale extraction condition is satisfied scale for each phrase based on the calculated satisfied degree, each according to the first extracting means
Based on the result of extracting the scale for each phrase,
Detected phrases with multiple scales
A plurality of continuous frames from the phrases
Second extraction means for extracting a plurality of mutually different scales for a plurality of continuous phrases by detecting a plurality of phrases, and for each scale among the plurality of scales extracted by the second extraction means, Summing means for summing the degrees of satisfaction calculated by the calculation means for a plurality of phrases to be executed, and a scale having a large sum result by the summation means among the plurality of scales extracted by the second extraction means. Scale selecting means for selecting as a scale of a plurality of phrases.

A third music information analyzing apparatus according to the present invention comprises: calculating means for calculating a degree by comparing temporally preceding and succeeding pitches for each of a plurality of phrases for which a certain scale is determined ; Determining means for determining the key of the section of the plurality of phrases based on the calculation result of the calculating means and the scale.

[0013]

According to the first music information analyzing apparatus described above, a plurality of frames are present between a plurality of phrases having the same scale.
Of the phrases for which the same scale as the
Search for phrases that meet the specified scale continuation conditions
And the searched phrase is the same as the surrounding phrase
Scaled . Therefore, the scale flow is
Or, even if it is interrupted over a plurality of phrases, the flow of the scale becomes clear, and it becomes possible to cope with an ambiguous scale.

Further, according to the second music information analyzing apparatus described above, a plurality of continuous phrases are different from each other.
When a plurality of scales are extracted by the second extracting means, a total
Means for each scale of multiple scales related to extraction
Is calculated by the calculating means for a plurality of phrases
Sum the degree of establishment. Then, the scale selecting means extracts
Out of multiple scales,
Selected as a scale for multiple consecutive phrases
You. Therefore, it is possible to determine the scale close to the human sense.
When multiple consecutive phrases exist in the middle of a song
Can respond to partial modulation .

In the above-mentioned first or second music information analyzing apparatus, there may be provided a phrase dividing means for classifying the music based on the average note length in the whole section or a partial section of the music. Good. This makes it possible to divide the phrase closer to the human sense, and thus to make the scale determination closer to the human sense.

According to the third music information analyzing apparatus described above, a plurality of phrases for which a certain scale is determined are calculated for each phrase by pitch comparison, and a plurality of phrases are calculated based on the calculation result and the scale. The key of the section is determined. Therefore, it is possible to perform a key detection close to a human sense for the natural scale.

In the third music information analyzing apparatus, the calculating means individually compares one pitch with a plurality of pitches temporally preceding the pitch to determine the long and short points, respectively. The length may be calculated by multiplying each of the length points by a coefficient determined so as to be larger as the pitch is closer to the one pitch. In this way, it is possible to calculate the degree of importance with an emphasis on the sense of length of a sound approaching in time, and it is possible to make the key determination closer to the human sense.

[0018]

FIG. 1 shows a music information analyzing apparatus according to an embodiment of the present invention. In this apparatus, music information processing is executed by a microcomputer.

An input device 12, a display 14, a central processing unit (CPU) 16, a program memory 18, a working memory 20, a melody memory 22, and the like are connected to the bus 10.

The input device 12 includes a keyboard and various input switches. By operating the keyboard in the input device 12, music information (information such as pitch, note length, rest length, etc.) corresponding to the melody of the desired music can be written into the memory 22. Music information can also be written into the memory 22 from another musical instrument or the like via the input device 12.

The display 14 displays a musical score or the like based on the music information stored in the memory 22.

The CPU 16 executes various processes such as taking in music information, dividing a phrase, detecting a scale, detecting a key, displaying a musical score, and the like in accordance with a program stored in a memory 18 composed of a ROM (read only memory). These processes will be described later with reference to FIGS.

The working memory 20 is composed of a RAM (random access memory), and includes a storage area used as a register or the like when the CPU 16 performs various processes. Registers related to the embodiment of the present invention include a phrase number register J, phase number registers K and L, key code registers KKC and LKC, note name registers KN and LN, and magnitude registers X, SUM, and S. These registers are used in the tone detection processing shown in FIGS.

The melody memory 22 is composed of a RAM and stores music information input from the input device 12 as described above. The memory 22 also stores phrase delimiter information set for music information, key information detected based on music information, and the like.

FIG. 2 shows the flow of processing of the main routine. In step 30, melody input processing is performed, and melody information of a desired song is input as music information from the input device 12 and written into the memory 22. Then, based on the melody information stored in the memory 22, for example, FIG.
A musical score as shown in FIG.

Next, at step 32, a phrase division process is performed. This processing will be described later with reference to FIG. Then, the process proceeds to a step S34.

In step 34, a scale detection process is performed. This processing will be described later with reference to FIG.
As a result of the scale detection processing, scale information representing a scale (for example, a G scale) related to the input melody information is obtained. After step 34, step 36
Move on to

In step 36, a key detection process is performed.
This processing will be described later with reference to FIGS. As a result of the key detection processing, key information representing a key (for example, when a G scale is detected, either a G major or an E minor) related to the input melody information is obtained.

FIG. 3 shows a subroutine of the phrase division processing. The process of FIG. 3 is performed on the melody information stored in the memory 22 as shown in FIG.

First, at step 40, a rest length adding process is performed. In this processing, a rest located next to a note is searched one by one, and every time the rest is searched, the length information of the rest is added to the length information of the note. . This is a process that takes into account that when a certain sound is played and a rest occurs, the last sound remains in the listener's ear in a deep color. As a result, the melody information can be processed as two types of information, pitch information and note length information.

As an example, when the quarter rest R is searched in the melody information of FIG. 4A, the length of the rest R is added to the length of the preceding quarter note, and the length of the rest R is added. As shown in B), the note before the rest R is a half note.

After the rest length addition processing, the melody information as shown in FIG. Then, the process proceeds to a step 42.

In step 42, an average note length (average note length) is calculated based on note length information of all sections of the melody. For the sake of simplicity, an example in which the average note length of the section shown in FIG. 4B is obtained instead of the entire section is shown in Table 1 below. The note length is represented by the number of points, with the 32nd note as one point.

[0034]

[Table 1] Next, the average note length a is multiplied by a constant b to obtain a border note length that matches or approximates the product ab, and the end of the border note length is used as a phrase delimiter, and a phrase number is assigned to each phrase. In the example of FIG. 4B, the average note length a =
Multiplying 7.65 by b = 2 results in ab = 15.3, and the border note length is a half note length (P = 16). Therefore, as shown in FIG. 4C, a phrase delimiter is set corresponding to the end of the half note length, and a phrase number is assigned to each phrase.

Thereafter, the melody information obtained by dividing the phrase is displayed on the display 14 together with the phrase number as shown in FIG. Then, the process proceeds to a step 44.

At step 44, a phrase division correction process is performed to correspond to the "slow portion". In other words, the processing in step 42 makes it possible to divide a phrase suitable for a song and close to the human sense. However, since the average note length of the entire melody is used, the “fast portion” and “slow portion” in one song are In any case, it is not easy to obtain a feeling of phrase close to humans. However, if the average note length is determined for each section by dividing it into multiple sections without applying the melody, the average note length changes depending on how the sections are divided. Therefore, it is preferable that one music piece (entire melody) has one average note length. Therefore, in this embodiment,
The phrase division in step 42 is performed by fineness corresponding to the "early part", and the "slow part" is handled in step 44.

In step 44, the average note length is calculated for each phrase based on the note length information in the phrase. In this case, the method of calculating the average note length is the same as that shown in Table 1. Then, it is determined whether the following two expressions are satisfied with respect to each of the two successive phrases, and if the result of this determination is negative, the break between the two phrases is deleted.

[0038]

## EQU00001 ## Average note length of phrase.times.2.ltoreq.note length of phrase For the phrases 1 to 5 shown in FIG. 4C, the average note length and the end note length of the phrase are shown for each phrase.
Table 2 below.

[0039]

[Table 2] In Table 2, the average note length and the terminal note length are both represented by the number of points as in Table 1.

The last note length of phrase 1 is 24,
The average note lengths of phrases 1 and 2 are 8 and 16 respectively.
It is. A constant multiple (double) of the average note length is 16 for phrase 1 and 32 for phrase 2. The last note length 24 of phrase 1 is twice (16) the average note length of phrase 1
Since it is larger, the condition of Expression 1 is satisfied. Therefore,
The break between phrases 1 and 2 is left as is.

Next, the last note length of phrase 2 is 16, and the average note length of both phrases 2 and 3 is 16
It is. Twice the average note length is 32 for phrases 2 and 3 respectively. Since the last note length 16 of the phrase 2 is smaller than the average note length 32 of any of the phrases 2 and 3, the condition of Formula 1 is not satisfied. Therefore, phrases 2 and 3
The separator between is removed. The result is shown in FIG.

Thereafter, the same processing as described above for phrases 1 to 3 is performed until the end of the music. As a result, the phrase division is corrected so as to conform to the “slow section” until the end of the song. After the musical score as shown in FIG. 5A including the correction result is displayed on the display 14, the process proceeds to step 46.

In step 46, a phrase division correction process is performed in order to bring the phrase closer to that of a human. That is, while listening to the melody, a human can recognize the same melody part as a phrase, but the program up to step 44 cannot divide the same melody part into phrases. For example, if the same melody part exists in two places, one with a long note immediately before it and the other with a note that is not very long, the program will recognize one melody part as a phrase, May not be accepted as a phrase (accepted as part of the previous phrase).
However, humans have the ability to have the melody feel the same parts as phrases. Therefore, in this embodiment, the same melody part is detected in steps 46 and 48 to perform phrase division.

In step 46, another phrase including the same melody portion as a certain phrase is detected, and it is determined whether the remaining portion of the detected phrase other than the same melody portion is longer than the minimum length of the phrase (for example, two beats). If the result of this determination is affirmative, the same melody part and the remaining part are divided as phrases. If the result of the determination is negative, it is determined that the image is overacted, and the image is not divided. To determine the melody identity between phrases, a pattern matching method that detects a pitch match for each unit length (for example, eighth note length) in the time axis direction can be used.

As an example, in the melody of FIG.
Phrase 4 contains the same melody part as phrase 1. In step 46, after the phrase 4 is detected, it is determined whether the remaining portion other than the same melody portion has two or more beats, and the determination result is affirmative. Therefore, the same melody part and the remainder are divided as phrases 4 and 5, respectively, as shown in FIG. After displaying the musical score including such a correction result on the display 14, the process proceeds to step 48.

In step 48, measures having the same melody in each phrase are detected, and the detected measures having the same melody are divided into phrases. For example, when there is a phrase as shown in FIG. 6A, the first, second, and third measures are determined to be the same melody by the pattern matching method,
As shown in FIG. 6 (B), each measure is composed of phrases 1, 2, 2
It is divided as three. After displaying the musical score including such a correction result on the display 14, the process returns to the routine of FIG.

FIG. 7 shows a subroutine of the scale detection process. In step 50, the appearance frequency of the pitch name is calculated for each phrase. The frequency of appearance of note names is 32 for each note, such as a single note and a tuplet, that constitute the melody.
Divided by diaeresis length unit determines the number of points of the sound mark as one point for each note 32 minutes, further sums the number of points of a plurality of notes sound name the same (e.g. C ₃ and C ₄₎ in the phrase It is calculated by: For example, FIG.
In phrase 1 of (A), the eighth note of note name C is 1
Since it appears only once, the appearance frequency of the pitch name C is 4 in the number of points.

As described above, the phrase 1 in FIG.
In Table 2, when the appearance frequency is calculated for each of the 12 pitch names C to B, the frequency is as shown in Table 3 below.

[0049]

[Table 3] Next, in step 52, the appearance rate of the pitch name is calculated for each phrase. The appearance rate n of the pitch name is obtained by an expression of n = m / k × 100, where m is the frequency of occurrence of the pitch name and k is the phrase length. Table 3 shows the phrase length in points for each phrase. In the example of Table 3, the appearance rate of each note name for each phrase is obtained as shown in Table 4 below.

[0050]

[Table 4] In this case, the total of the appearance rates of the pitch names C and B for each phrase is about 100, excluding the error of the minor part.

Next, at step 54, C
Calculate the degree of establishment of a scale based on the pitch names of B to. The establishment degree of the scale is obtained by summing up the appearance degrees of the pitch names that constitute the scale. For example, in the case of the C scale, the appearance rates of note names C, D, E, F, G, A, and B may be summed. Table 5 below shows the result of obtaining the degree of establishment of the scale in the example of Table 4.

[0052]

[Table 5] The degree of establishment of the scale increases as the sum of the appearance rates of the scale-constituting sounds increases. In the example of Table 5, phrase 1
Is the G scale and phrase 2 is the C scale.

When calculating the degree of establishment of the scale, the phrase division is important. That is, a method has been proposed in which a bar delimiter is used as a phrase delimiter, but this has a problem. For example, in the example of FIG. 8A, if the first two sounds are included in the phrase 1 by matching the phrase 2 to the bar line, the sound ratio changes. If this is the part of transposition / transposition, there is a possibility that the decisive sound may be lost, and this is not a good method. Even in the example of FIG. 8A, if the first two sounds of phrase 2 disappear,
Since the note name B is lost, it is difficult to distinguish between the C scale and the F scale.

In this embodiment, the phrase division close to the human sense is performed by the processing described with reference to FIG. 3, and the scale detection is performed based on the result. Close scale detection is possible.

Table 6 below exemplifies the result of calculating the degree of establishment of the scale of C to B for each phrase in a song (melody) divided into 20 phrases.

[0056]

[Table 6] In step 56, a scale satisfying a predetermined scale extraction condition is extracted for each phrase. The scale extraction condition is
As an example, it is assumed that a scale with a degree of establishment of 100% is preferentially extracted, and a scale with a degree of establishment of 90% or more is extracted if there is no 100%. Table 7 below shows the result of extracting the scale from Table 6 according to the scale extraction conditions.

[0057]

[Table 7] In the case of a song that has no transposition and does not smell any other key (melody), there will be no sound outside the scale. In such a case, if a table such as Table 7 is created, 100% of columns appear on all the phrases at a certain scale. However, in the case of Table 7, one scale did not appear over the entire phrase because sounds other than the specific scale appeared. This means that transposition / transposition is possible.

From the arrangement of scales as shown in Table 7, the arrangement of scales that humans would judge is to be found. It turns out to be strong. Also, there is a possibility that A # may occur in the middle, and there is a high possibility that the last will be either C or G.

By the way, there is a method of determining the scale by looking at the arrangement of the scale in real time. In this method, G up to phrase 2 and A # from phrase 3
Is more likely to choose. However, when considering non-real time, it is understood that there is a possibility that phrase 4 will return to G. Therefore, in the real-time determination method that determines the scale in ascending order of the phrase number,
It turns out that it does not match the non-real-time human sense.

In Table 7, for example, when the length of phrase 3 is short, the tone of a person becomes ambiguous for a moment,
I can imagine returning to G scale again in phrase 4. If you are aware of the flow of the scale, the impression of transposition / transposition will diminish the moment it returns, even if a different key appears momentarily in a short section. However, if a different key strongly appears and continues over a long section, you will feel the transposition / transposition.

Therefore, in step 58, a scale continuation process for continuing the break of the scale is performed. That is, one or a plurality of phrases that satisfies a predetermined scale continuation condition among phrases in which the same scale is interposed between a plurality of extracted phrases and in which the same scale as the plurality of phrases is not extracted. A phrase is searched, and the searched phrase has the same scale as the preceding and following phrases. As an example of the scale continuation condition, even if the length of the phrase of the same scale not extracted is within 4 bars or the length of the phrase is longer than 4 bars, the average of the degree of scale establishment in the phrase is 90%. We will continue to scale beyond what we have. Table 8 below shows the result of performing the scale continuation process in the example of Table 7 and setting “−1” to the continuation section.

[0062]

[Table 8] According to Table 8, phrases 3 and 7 are regarded as G-scale continuation sections, and phrases 1 to 11 are a series of G-scale sections. In this example,
None of the sections longer than four measures had an average of the degree of scale formation exceeding 90%.

Next, in step 60, one scale that continues from the beginning to the end of the music is extracted, and if there is no such scale, a plurality of scales that connect from the beginning to the end of the music are extracted. In the example of Table 8, there is no single scale that continues from the beginning to the end of the song, so when multiple scales that connect from the beginning to the end of the song are extracted,
The results are as shown in Table 9 below.

[0064]

[Table 9] Next, in step 62, it is determined whether there is a phrase in which a plurality of scales overlap. If the result of this determination is negative (N), the flow proceeds to step 64, and the extracted scale is determined as the scale of the extraction section. Then, scale information representing the determined scale is stored in the memory 20 and displayed on the display 14.

In step 60, if one scale that continues from the beginning to the end of the song is extracted, or if a plurality of scales that connect from the beginning to the end of the song are extracted without overlapping, the steps from step 62 64
And the extracted one or more scales are used as the scale of the extraction section. After step 64, the process returns to the routine of FIG.

If the decision result in the step 62 is affirmative (Y), there is a phrase in which a plurality of scales overlap.
That the multiple scales that are different from each other
Is detected), and the routine proceeds to step 66.
In step 66, of the overlapped scales, the one having a larger sum of the scale establishment degrees in the phrase is extracted as the phrase scale. That is, it is detected in step 62
Phrase (multiple different scales were extracted
Multiple consecutive phrases from among phrases)
Then, based on this detection result and the detection result of step 62,
Multiple different phrases for multiple consecutive phrases
Extract the scale. Then, multiple scales related to extraction
Multiple consecutive phrases for each scale of the
And the sum of the degrees of scale formation is calculated.
Extracted as a scale of multiple consecutive phrases (selection
Select). In the example of Table 9, the A # scale and the C and G scales overlap in the 16th and 17th phrases, and the total scale establishment degree is 190 for A # and 1 for C and G, respectively.
Since it is 95, the C and G scales are extracted for the 16th and 17th phrases.

Next, in step 67, it is determined whether or not there are overlapping scales and the sum of the degrees of scale establishment in the phrase is equal. If the result of this determination is negative (N), the operation moves to step 64, and the scale is determined in the same manner as described above. For example, assuming that in the case of Table 9, the C scale is not extracted in step 60 and the G scale is extracted for the 16th and 17th phrases in steps 62 and 66 ,
The 1st to 11th phrases have a G scale, the 12th to 15th phrases have an A # scale, and the 16th to 20th phrases have a G scale.

If the result of the determination at the step 67 is affirmative (Y), the routine goes to a step 68. In step 68, the sum of the appearance rates of the first (basic), third, fifth, and sixth sounds is calculated for the overlapped scales, and the scale with the highest sum is used as the scale of the phrase having the scale overlap. decide. Then, scale information representing the determined scale is stored in the memory 20 and displayed on the display 14. Thereafter, the process returns to the routine of FIG.

In the example of Table 9, the C and G scales are 16th.
Since the で 20 to 重 20 phrases are overlapped and the sum of the scale establishment degrees in the phrase is equal to 485, the scale of the 第 16２０20th phrase is determined by the processing of step 68. In this case, the sum of the appearance rates of the 1, 3, 5, and 6th sounds is G
If the C scale is larger than the scale,
The 6 to 20 phrases are on the C scale, and the final scale flow of the song is as shown in Table 10 below.

[0070]

[Table 10] According to Table 10, the 1st to 11th phrases are on the G scale, the 12th to 15th phrases are on the A # scale, and the 16th to 20th phrases are on the C scale.

In the state shown in Table 10, the scale detection has been completed, and the key cannot be determined for the natural scale. For example, even if it is determined that the first to eleventh phrases are on the G scale, it is unknown whether the key is G major or E minor. Therefore, key detection processing is performed next based on the scale detection result.

FIG. 8 (B) shows one of the tones to be detected.
It shows two phrases. When the melody moves like this phrase, the human perceives the length of the phrase as a whole based on the length (measure / minor feeling) of each moment when the sound is heard. Further, the sense of length at the moment when a certain sound is heard is not determined only by the immediately preceding sound, but is also affected by sounds before that sound.

In order to simplify the processing procedure, FIG.
Is divided into phases of unit length, and the pitch is detected for each phase. Here, since the shortest note is an eighth note, the unit length is set to an eighth note length. However, the unit length is matched with a division unit (for example, a 32nd note length) in calculating the appearance frequency of the note name described above. Best.

FIG. 9A shows the pitch of the phrase shown in FIG. 8B by the key code value for each eighth note length phase. FIG. 9B is for explaining a method of calculating the length in a certain phase Fi in FIG. 9A. To calculate the magnitude of the moment the sound is heard,
It is necessary to clarify the causal relationship between the change in pitch and the sense of length. Therefore, as shown in Table 11 below, a long / short point table showing correspondence between a pitch change between two tones and a long / short feeling by a long / short point of 0, +1 or -1 was created. This table is for the C scale (major), and for other scales, it is necessary to adjust whether to calculate as C major based on the scale information or to adjust the data in Table 11 to the scale.

[0075]

[Table 11] Here, “0” indicates that there is no sense of majority, “+1” indicates that there is majority, and “−1” indicates that there is minority. Such a table is stored in the memory 18 of FIG.

When comparing two tones using Table 11, a pitch name with a high pitch is specified in a horizontal row, and a pitch name with a low pitch is specified in a vertical column. For example, in comparison of pitch names G and E, when G is high like G ₃ and E ₃ , the long and short point is “+1”, and when E is high like G ₃ and E ₄ , the long and short point is “−”. 1 ".

In order to calculate the length of a certain phase, it is desirable to consider not only the sound of the phase immediately before the phase but also the sounds of a plurality of previous phases. Therefore, in the example of FIG. 9B, the sound in one measure (eight phases) before phase Fi is considered, and the pitch of Fi is compared with the pitch of each phase in one measure before that. Table 11
The long and short points R were determined according to the following.

Also, even the sound within one bar before the phase Fi has a greater influence as the sound is closer to Fi.
As shown in (B), the coefficient S = 8 to 1 is determined so that the coefficient becomes larger as the value is closer to Fi. Then, the product (R × S) is obtained by multiplying the long and short point R by the coefficient S for each phase, and the products for the eight phases thus obtained are summed to obtain the phase length of Fi. In the example of FIG. 9B, the phase length is +8, and it can be seen that the sense of major key is strong.

When the above-described one-phase length calculation processing is applied to each phase in one phrase, a phase length is obtained for each phase as shown in Table 12 below. However, the numerical values in Table 12 are not actually calculated,
It is inserted for convenience of explanation.

[0080]

[Table 12] Next, the length of the phrase (phrase length) is determined.
The phrase length is calculated by summing the lengths of all phases in one phrase. In the example of Table 12, the phrase length is +78, indicating that the phrase is in a major tone.

The phrase length is calculated for each phrase from the beginning to the end of one song by the above method. Table 1 below
Numeral 3 indicates the result of scale detection and the phrase length for each of the songs divided into 15 phrases.

[0082]

[Table 13] Thereafter, the length (section length) of a section composed of a plurality of phrases of a certain scale is obtained, and the key of the section is determined based on the obtained section length and scale. In the example of Table 13, the section length of the first to tenth phrases on the C scale is +315, and the section length of the first to fifteenth phrases on the G scale is -156. Accordingly, the key of the section of the first to tenth phrases is C measure, and the key of the section of the first to fifteenth phrases is E minor.

FIGS. 10 and 11 show a subroutine of a key detection process for enabling the above key detection.
In step 70, 1 is set in the phrase number register J. This is to calculate the degree from phrase 1.

Next, at step 72, 2 is set to the phase number register K and the length register SUM is set.
Is set to 0. Then, the process proceeds to a step 74, wherein a value obtained by subtracting 1 from the value of the register K is set in the register L. As a result, when the process first reaches step 74 after setting K = 2 in step 72, L = 1, and the pitches of phases 2 and 1 can be compared. After step 74, the process moves to step 75.

In step 75, it is determined whether the value of the register L is 0. If the result of this determination is negative (N), the operation moves to step 76, where the key codes of the K and L-th phases are set in registers KKC and LKC, respectively. Here, the K-th and L-th phases are specified by the values of the registers K and L, respectively. Step 7
When step 76 is reached for the first time after 2, 74, the key codes of the second and first phases are set in KKC and LKC.

Next, at step 78, a pitch name detection process is performed. That is, the value of the register KKC is divided by 12 (integer operation) to obtain a remainder (KKC mod
The surplus value corresponding to the pitch name is set in the register KN according to 12). As an example, pitches C ₄ , C ₄ #... E are represented by key code values of 60, 61... 64..., Pitch names C, C #.
When the operation of d12 is performed, 4 corresponding to the pitch name E is set in KN. Similarly, the operation of LKC mod 12 is performed on the register LKC, and the remaining value corresponding to the pitch name is set in the register LN. Then, the process proceeds to step 80.

At step 80, the registers KKC, LK
Pitch relation determined by C key code and registers KN, LN
The long / short point R is read from the long / short point table (Table 11) in the memory 18 on the basis of the note name information. Since the memory 18 also stores the coefficient S, the coefficient S corresponding to the phase proximity determined by the difference between the values of the registers K and L is read from the memory 18. Then, the read (long and short) point R is multiplied by the read coefficient S, and the product (R × S) is set in the degree register X. The value of X at this time is the phase length. As an example, as shown in FIG.
8 when R = + 1 and S = 7 are read, R × S
= + 7 is set to X.

Next, at step 82, the value of the register X is added to the value of the degree register SUM, and the addition result is set in SUM. Then, the process proceeds to step 84.

In step 84, it is determined whether or not the value obtained by subtracting the value of the register L from the value of the register K is 8 (whether the value has reached eight phases before). When the process first arrives at the step 84 after the steps 72 and 74, since K = 2 and L = 1, the determination result is negative (N), and the process proceeds to the step 86.

At step 86, the value of the register L is decreased by one. Then, returning to step 75, it is determined whether L = 0. For the first time after steps 72 and 74, step 86
, Since L = 0, the determination result of step 75 is affirmative (Y), and the routine goes to step 88.

At step 88, it is determined whether the K-th phase is the last phase of the phrase. Steps 72 and 74
The first time after that comes to step 88, K = 2,
Since it is not the final phase, the determination result is negative (N), and the routine proceeds to step 90.

At step 90, the value of the register K is increased by one. Then, the process returns to step 74, and the value K-1 is set in the register L. As a result, it is possible to calculate the degree in the next phase. For example, step 90
If K = 3 in step 74, L = 2 in step 74, and the pitches of phases 3 and 2 can be compared.

As described above, the processing of steps 74 to 88 is performed until the value of the register K becomes 9 in step 90. In the processing so far, KL = 8 is not satisfied, and the process proceeds from step 75 to step 88 when L = 0.

On the other hand, when K = 9 in step 90, 8 is set in the register L in step 74, and the phase 9
And 8 can be compared. And step 75
Steps 86 through 86 are performed until L = 1 in step 86.

When L = 1 at step 86, the processing of steps 76 to 82 is performed via step 75. Then, at step 84, the determination result is affirmative (Y) because KL = 9-1 = 8, and the routine goes to step 88. In this case, the degree of the phase 9 is calculated in consideration of the pitches of the eight phases before the phase 9 in the same manner as described with reference to FIG. 9B.

If the phase 9 is not the final phase of the phrase 1, the process goes from the step 88 to the step 90,
Assume that the value of the register K is 10. Then, after setting 9 to the register L in step 74, the processing of steps 76 to 86 is performed eight times through step 75.
Is affirmative (Y), and the routine goes to Step 88. Therefore, after K = 9 in step 90, the phase length and length are calculated for each phase in consideration of the pitches of the previous eight phases, as described with reference to FIG. 9B. become.

Thereafter, when the K-th phase reaches the final phase of phrase 1 while repeating the processing of steps 74 to 90, the determination result of step 88 becomes affirmative (Y), and the routine proceeds to step 92.

At step 92, the value of the register SUM is stored in the memory 20 as the phrase length. As a result, when the final phase of phrase 1 has been reached as described above, the magnitude of phrase 1 is stored in memory 20. Thereafter, the process proceeds to step 94.

At step 94, it is determined whether the J-th phrase is the last phrase in the music. The Jth phrase is
It is specified by the value of the register J. When the process first proceeds to step 94 after setting J = 1 in step 70, the determination result in step 94 is negative (N), and the process proceeds to step 96.

At step 96, the value of the register J is increased by one. Then, the process returns to step 72, and the subsequent processing is repeated in the same manner as described above. As a result, when J = 2 in step 96, the processing of steps 72 to 94 is performed on phrase 2, and
The same processing is performed for 4.

When the J-th phrase reaches the last phrase as a result of such processing, the determination result of step 94 becomes affirmative (Y), and the routine proceeds to step 98 of FIG. At this time, the phrase lengths of all the phrases of the music are stored in the memory 20, and the stored information is displayed on the display 14.

The processing after step 98 is to determine the key based on the scale information and the phrase length stored in the memory 20. In step 98, 1 is set in the phrase number register J and 0 is set in the degree register S. Then, the process proceeds to step 100.

In step 100, the value of the register S is
The value obtained by adding the length of the second phrase is set in S. Then, the process proceeds to step 102.

At step 102, the J-th phrase is the last phrase in the music or the J-th phrase and J + 1
It is determined whether the scale is different from the second phrase. When it comes to step 102 for the first time after step 98, J =
1 and not the last phrase. Table 1 above
Assuming that the phrases 1 and 2 have the same scale as shown in FIG. 3, the determination result of step 102 is negative (N).
Then, the process proceeds to step 104.

In step 104, the value of the register J is set to 1
Only increase. Then, the process returns to step 100 to perform the above-described addition processing of the degree. If J = 1, J = 2 in step 104, and step 10
At 0, the phrase lengths of the phrases 1 and 2 are added and set in the register S.

When the J-th phrase reaches the last phrase while repeating such processing, step 1
02 is affirmative (Y), and step 106
Move on to Also, as shown in Table 13 above, the phrase 1
When the scale is different between 0 and 11, step 10
Move to 6.

At step 106, the value of the register S is 0
It is determined whether or not the above. At this time, the total value (section length) of the first to Jth phrases is set in S. If the determination result of step 106 is affirmative (Y), the process proceeds to step 108, where the key of the performance section including the first to J-th phrases is determined so that the main tone is the same as the fundamental tone of the scale. For example, as shown in Table 13 above, when the scale is C in the phrases 1 to 10 and the section length is +315, the key of the section of the phrases 1 to 10 is determined as the C measure.

If the result of the determination in step 106 is negative (N)
If so, the process proceeds to step 110. Step 11
At 0, the key of the performance section consisting of the first to J-th phrases is determined as the minor key whose main tone is 6 degrees above the fundamental of the scale. For example, in the above-described C scale phrases 1 to 10, when the section length is a negative value, the key of the section of the phrases 1 to 10 is determined to be A minor.

When the processing in step 108 or 110 is completed, the process proceeds to step 112. In step 112,
It is determined whether the J-th phrase is the last phrase in the song.
As described above, when it is the last phrase in the determination in step 102, the determination result in step 112 is also affirmative (Y), and the process returns to the routine in FIG.

On the other hand, if the scales of the successive phrases are different in the determination in step 102, the process proceeds to step 112.
Is negative (N), and the routine proceeds to step 114. In step 114, 0 is set in the register S. Then, in step 104, the value of the register J is incremented by 1, and the process returns to step 100, and the subsequent processing is repeated in the same manner as described above.

As an example, in Table 13 described above, when the section of phrases 1 to 10 is determined to be the C measure by the processing of step 108, step 112 to step 1
The process proceeds to step 104 via 14, and J = 11. Then, step 1 is performed until J = 15 in step 104.
The processes of 00 and 102 are repeated.

After the processing of step 100 is performed with J = 15 at step 104, when step 102 is reached, J = 15 is the last phrase, so the determination result is affirmative (Y), Move to step 106.

If it is determined in step 106 that the value of the register S is -156 as shown in Table 13, the determination result is negative (N), and the flow proceeds to step 110. In step 110, the scale of phrases 11 to 15 is G
Therefore, the key of the section of the phrases 11 to 15 is determined to be E minor.

The key information representing the key determined in steps 108 and 110 is stored together with the melody information in the memory 22 and displayed on the display 14.

The present invention is not limited to the above embodiment, but can be implemented in various modifications. For example,
The following changes (1) to (6) are possible.

(1) The constant such as 90% shown in the embodiment is
Other numbers may be used.

(2) The processing procedure described in the embodiment may be omitted as appropriate when accuracy is not important.

(3) In the embodiment, the scale detection relating to the natural scale has been described. However, the scale detection technique of the present invention can be applied to the detection of a melodic minor scale, a harmony minor scale, and the like.

(4) When calculating the length of a certain phase, the pitch is compared between the phase and the preceding eight phases. However, the pitch may be compared with a phase less or more than eight.

(5) In the routine of FIG. 10, the pitch comparison is performed for each phase until the phase becomes 7 or less until K = 9, but even before K = 9, each phrase is compared. Each time, the pitch may be compared with the preceding eight phases.

(6) The phrase division before the scale detection processing may be performed manually. Further, the scale determination before the tone detection processing may be performed manually.

[0122]

As described above, according to the present invention, the interruption of the scale can be continued or one of the overlapping scales can be selected so as to cope with the ambiguity of the scale and the partial modulation. The effect of enabling scale detection close to the sense of is obtained.

Further, for a plurality of phrases for which a certain scale is determined , the pitch is calculated for each phrase by comparing the pitches, and the key of the plurality of phrase sections is determined based on the calculation result and the scale. There is also obtained an effect that the key detection close to the human sense can be detected with respect to the natural scale.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a music information analyzer according to one embodiment of the present invention.

FIG. 2 is a flowchart showing a main routine.

FIG. 3 is a flowchart showing a subroutine of a phrase division process.

FIG. 4 is a diagram showing a musical score for explaining a processing example of steps 40 and 42.

FIG. 5 is a diagram showing a musical score for explaining a processing example of steps 44 and 46.

FIG. 6 is a diagram showing a musical score for explaining a processing example of step 48;

FIG. 7 is a flowchart illustrating a subroutine of a scale detection process.

FIG. 8 is a diagram showing a musical score of a phrase used in scale detection processing and key detection processing.

FIG. 9 is a diagram for explaining a method of calculating the length of a certain phase.

FIG. 10 is a flowchart showing a part of a subroutine of a key detection process.

FIG. 11 is a flowchart showing the remaining part of the subroutine of the key detection processing.

[Explanation of symbols]

10: bus, 12: input device, 14: display, 1
6: CPU, 18: program memory, 20: working memory, 22: melody memory.

──────────────────────────────────────────────────の Continued on the front page (58) Fields surveyed (Int.Cl. ⁷ , DB name) G10H 1/00 101-102 G10H 1/36-1/42 G10G 1/00-3/04

Claims (5)

(57) [Claims] 1. A calculating means for calculating a degree of establishment of a plurality of scales for each phrase based on pitch information of a song divided into phrases, and a predetermined value for each phrase based on the degree of formation calculated by the calculating means. Extracting means for extracting a scale that satisfies the scale extracting condition of the following, interposed between a plurality of phrases from which the same scale is extracted by the extracting means, and the same scale as the plurality of phrases is not extracted by the extracting means A music information analyzing apparatus comprising: a search unit that searches for a phrase that satisfies a predetermined scale continuation condition from the selected phrases, and sets the searched phrase to have the same scale as the plurality of phrases. 2. A calculating means for calculating a degree of establishment of a plurality of scales for each phrase based on pitch information of a song divided into phrases, and a predetermined value for each phrase based on the degree of formation calculated by the calculating means. Extracting means for extracting a scale satisfying the scale extracting condition of the following, and extracting the scale for each phrase by the first extracting means.
Multiple different scales are extracted based on the results
Of the detected phrase and the
A second extraction unit for extracting a plurality of different scales for a plurality of continuous phrases by detecting a plurality of continuous phrases from among the plurality of scales; and a plurality of scales extracted by the second extraction unit. Summing means for summing up the degrees of satisfaction calculated by the calculation means for the plurality of continuous phrases for each of the scales; and the summing means of the plurality of scales extracted by the second extraction means. A music information analyzing apparatus comprising: a scale selecting unit that selects a scale having a large total result as a scale of the plurality of continuous phrases. 3. The music information analyzing apparatus according to claim 1, further comprising a phrase dividing means for dividing a phrase of the music based on an average note length in an entire section or a partial section of the music. 4. A calculating means for calculating a degree by comparing the pitch before and after each phrase with respect to a plurality of phrases for which a certain scale is determined, and a calculation result of the calculating means and the scale. Determining means for determining a key of the section of the plurality of phrases based on the music information. 5. The calculating means individually compares one pitch with a plurality of pitches temporally preceding the pitch to determine a long-short point, and calculates a long-short point for each of the long-short points. 5. The music information analyzer according to claim 4, wherein the degree is calculated by multiplying each coefficient by a coefficient determined to be larger as the pitch is closer to the one pitch.