JPH0736478A - Calculating device for similarity between note sequences - Google Patents

Abstract

PURPOSE:To enable comparison between note sequences which effectively utilizes a clue that notes have. CONSTITUTION:In order to find the similarity between the note sequences, a note correspondence calculating means 3 finds the correspondence between the notes to be compared as to a clue note sequence and a candidate note sequence inputted by a clue note sequence input means 1 and a candidate note- sequence input means 2. On the basis of the correspondence of notes found by the note correspondence calculating means 3, a note matching degree calculating means 4 finds the degree of matching in note units and a note importance calculating means 5 finds the importance of corresponding notes. A similarity calculating means 6 finds the similarity between the note sequences from the degree of matching and importance of the corresponding notes and outputs the similarity to an output means 7.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a music retrieval system for retrieving music by using a musical note sequence as a clue, and a musical note sequence for calculating the degree of similarity between a clue musical note sequence and a candidate musical note sequence in a musical composition search. An inter-similarity calculation device.

[0002]

2. Description of the Related Art In a music search process using note string information as a clue, it is checked which musical piece contains a note string which is a clue to the search, and a search result is determined. In this process, in order to find out which song contains the note sequence that serves as the search clue, the note sequence between the clue note sequence and the candidate note sequence contained in the search target song is compared. is necessary. However, when the search clue is inaccurate, the clue note sequence and the candidate note sequence of the musical piece are not always exactly the same. In such a case, not only is it determined whether the clue note sequence and the candidate note sequence of the music match,
It is necessary to obtain a degree of similarity (hereinafter referred to as “similarity”) between the clue note string and the candidate note string of the music, and set the music having the maximum similarity as the search result.

Conventionally, in order to obtain the similarity between note sequences between a cue note sequence and a candidate note sequence of a musical composition, attention is paid to the notes included in the cue note sequence and the candidate note sequence, and the notes are compared from the note comparison results. A device for obtaining the similarity between columns is known. One example is "Melody Search-Humming to Music D"
Search B-, Proceedings of the 43rd National Convention of Information Processing Society of Japan, Information Processing Society of Japan, 3M-5, October 1991
Month (hereinafter referred to as Reference 1) ”.

The note string selecting device described in Document 1 will be described below with reference to the drawings. FIG. 11 is reference 1
FIG. 3 is a block diagram of a note string similarity calculation device described in FIG. With reference to FIGS. 4, 11 and 12A, the operation of the inter-note sequence similarity calculation device described in Document 1 will be described using a specific example. FIG. 4 is a diagram showing a specific example of a note string input to the note string similarity calculation device. In FIG. 4, the pitch value of the notes included in each note string is indicated by a value that the central C sound is 0 and is increased / decreased in semitone steps, and the pitch value is set to 1 for quarter notes and proportional to the pitch. Indicated by the value to

When calculating the similarity between note sequences of the cue note sequence 1 and the candidate note sequence 1 of FIG. 4, the cue note sequence input means 1 inputs the cue note sequence 1, and then the candidate note sequence input means 2 The candidate note string 1 is input. Next, the note correspondence calculation means 3 associates the clue note sequence 1 and the note of the candidate note sequence 1 in order from the beginning as the correspondence of the notes to be compared between the cue note sequence 1 and the note sequence of the candidate note sequence 1. The note coincidence calculation means 4 obtains a comparison result between notes based on the correspondence. For example, the pitch difference from the preceding note of each note is compared between the clue note string 1 and the candidate note string 1, and the pitch difference with the preceding note of the corresponding note is 1 and 0 if not. To determine the degree of note matching. The note string deformation point deduction calculation means 11 obtains a number obtained by multiplying the number of times a note string is inserted or deleted by -1 as a score deduction of the similarity according to the deformation of the note string. In the comparison between the clue note string 1 and the candidate note string 1, the note string to be compared is not modified, so the note string modification deduction point is set to 0. In the similarity calculation means 6, a value 3 which is the sum of the note coincidence and the value of the note string deformation deduction
Is output to the output means 6 as the similarity between note sequences.

[0006]

When a musical piece is searched by using a musical note sequence reproduced based on a user's memory as a clue, a part of the musical note sequence in which the memory is ambiguous tends to be inaccurate. The part of the note sequence that tends to be inaccurate is
Specifically, it is a fine portion where short notes continue and the beginning or end of a note string. For this reason, in order to effectively utilize the note string based on the user's memory as a clue for music retrieval, for example, a note with a long note length is emphasized to search for a match in the rough shape of the melody line. It is necessary to perform processing such as prioritizing as a clue for the search, and processing for emphasizing the middle portion as a clue for the search rather than the beginning and end of the note string.

However, in the conventional note string similarity calculating device described in Document 1, the note matching degree is normalized so that it becomes 1 when the corresponding notes are completely matched. Is obtained only by the total degree of note coincidence and the point deduction due to the deformation of the note sequence. In other words, in the inter-note-note-similarity calculating device described in Document 1, the similarity is calculated according to the proportion of correct notes in the entire note string. In other words, all the notes in the note sequence are treated equally regardless of the note length and the note position,
The importance of each note in the note string is not reflected in the similarity between the note strings. Therefore, there is a problem that an inaccurate portion of the note sequence is overestimated and an appropriate similarity between the note sequences cannot be calculated.

This problem will be specifically described with reference to FIGS. 4 and 12. In FIG. 4, the candidate note string 1 is different from the clue note string 1 in the middle part of the note string and in the pitch value of the note having a long note length indicating the skeleton of the melody. On the other hand, the candidate note string 2 is different from the clue note string only in the pitch value of the last short note. In this example, the note-string similarity between the candidate note string 2 and the clue note string 1 is higher than that between the candidate note string 1 and the clue note string 1 It is desirable to define it.

However, even in the example in which the skeleton note sequence 1 and the note sequence have different skeletons such as the candidate note sequence 1, only the clue note sequence 1 and the short note at the end of the note sequence differ in the candidate note sequence 2 Also in the above, there is no difference in the pitch difference value between the preceding note of one note and that of the clue note sequence 1. For this reason, when the note sequence similarity calculation device described in Document 1 obtains the note sequence similarity between the candidate note sequence 1 and the candidate note sequence 2 and the clue note sequence 1, it is calculated as 3 in all cases. . This is a problem that arises from the fact that the note sequence similarity calculation device described in Document 1 treats all the notes in the note sequence equally and cannot reflect the importance of each note.

The present invention solves the above-mentioned problems, and the purpose thereof is to calculate the similarity between musical note sequences that can reflect the importance of the notes of the musical note sequence included in the musical note sequence input as a clue. (EN) A device is provided, and as a result, it is possible to perform a music search effectively using a note sequence reproduced based on a user's memory as a clue.

[0011]

DISCLOSURE OF THE INVENTION The present invention provides a clue note string input means for inputting a cue note string in a note string similarity calculation device for comparing and collating a cue note string and a candidate note string. A candidate note string input means for inputting the candidate note string, a note correspondence calculation means for obtaining correspondence between the clue note string and the note to be compared between the candidate note string, and the clue note string and the candidate note based on the correspondence A note matching degree calculating means for obtaining a note matching degree in a note unit between strings, a note importance calculating means for obtaining a note importance degree in a note unit of the clue note string based on the correspondence, and the note matching degree based on the correspondence And a similarity calculation means for calculating the similarity between note strings from the note importance, and an output means for outputting the similarity between note strings.

[0012]

DESCRIPTION OF THE PREFERRED EMBODIMENTS An apparatus embodying the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of the present invention. Cue note string input means 1 for inputting a cue note string, candidate note string input means 2 for inputting a candidate note string, and note correspondence calculation for obtaining a correspondence relationship between a cue note string and a note to be compared between the candidate note string The means 3 and the note coincidence calculation means 4 for obtaining the degree of coincidence between the notes included in the clue note sequence and the candidate note sequence based on the correspondence between the notes.
And a note importance degree calculation means 5 for obtaining the importance degree of a clue note based on the correspondence relationship between notes, and a similarity degree calculation means for obtaining a similarity between note strings from the note coincidence degree and the note importance degree based on the correspondence relationship between notes 6 and output means 7 for outputting the similarity between note sequences.

The clue note string input means 1 is provided with a note string serving as a clue for music retrieval. As the input means, input using a keyboard such as a keyboard instrument, note input using a character keyboard, input using MIDI which is a musical instrument control data transfer standard, input using a mouse / touch panel / trackball / pen input device, Input from a storage device, input by generating a note sequence from an acoustic signal obtained by a microphone or the like (for example, "Personal computer music system-automatic transcription of singing voice-, Proceedings of the 35th IPSJ National Convention Proceedings, Foundation" Information Processing Society of Japan, 5
Ff-5, 1987 ”, which uses an automatic transcription device) and the like. The input note sequence is a sequence in which multiple notes are arranged. The note string may include rests at the beginning, between notes, between notes, and the like. Also,
Each note has, as an attribute value, one or more pitch value representing the pitch of each note and a note length value representing the note length. Each rest may have a rest length attribute value indicating the length of the rest. Furthermore, each note may have a note reliability that indicates the degree of accuracy of the attribute value of the note.

The candidate note string input means 2 is supplied with a note string which is a candidate for music retrieval. As input means, input using a keyboard such as a keyboard instrument, note input using a character keyboard, input using MIDI which is a musical instrument control data transfer standard, input using a mouse / touch panel / trackball / pen input device, Input from storage device,
A note string can be generated from an acoustic signal obtained by a microphone or the like to be used as input. The input note sequence is a sequence in which multiple notes are arranged.

The note correspondence calculation means 3 receives the clue note string from the clue note string input means 1 and the candidate note string input means 2
Each candidate note sequence is received from, and the correspondence of the notes to be compared is calculated between the received note sequences. As an example of how to find the correspondence between notes, when i is a natural number, the correspondence between the i-th note from the beginning of the note string, the correspondence between the i-th note from the end of the note string, and the note string It is possible to use the one corresponding to the i-th note or the rest from the beginning of the above, or the one corresponding to the i-th note or the rest from the end of the note string.

The note coincidence calculation means 4 corresponds to a note for comparing a clue note string from the clue note string input means 1, a candidate note string from the candidate note string input means 2 and a note correspondence string from the note correspondence calculation means 3. The relation is received, and the note coincidence degree representing the degree of coincidence of the corresponding notes is obtained based on the correspondence relation. The note matching degree can be determined by, for example, matching of pitch values, matching of pitch values, matching of pitch values and pitch values, and the like. In addition, the pitch difference between the corresponding note and the immediately preceding note in each note string is compared, and the value is determined from the difference in pitch difference, mismatch in pitch ratio, and mismatch in pitch difference and pitch ratio. Can also be defined. When the note string includes rests, whether or not the rests are supported, whether the rest lengths of the rests match or not, and the like can also be used as the note coincidence value. Further, regarding the coincidence of these elements, not only the value is changed depending on whether they are completely coincident or not, but also the degree of coincidence of note can be obtained according to the degree of coincidence.

The note importance calculation means 5 calculates the clue note string from the clue note string input means 1 and the note correspondence calculation means 3
From the note sequence corresponding to the note to be compared with each other, and the note importance degree is calculated for each note included in the clue note sequence based on the correspondence. The note importance can be, for example, a note length value of a note in the cue note sequence,
It can be decided that the note importance becomes higher as the note length becomes longer. Also, not only the note length of the note, but the sum or product of the note length of the note and the note length of the immediately preceding note.
It can be set to a value such that the note importance increases as the note length of one note increases. Further, when the note sequence includes rests, the note importance can be increased as the rest length becomes longer, such as a value proportional to the rest length. Although these have decided the note importance degree based on the note length of a note or the rest length of a rest, the note importance degree can also be decided based on the position information in the note string of a note. For example, as the note importance, as shown in FIG. 2, a low value is given at the beginning and the end of the note sequence, and a high value is given at the middle of the note sequence. In this way, note importance can be determined by the note length, rest rest length, note or rest position, etc., but note importance depends on the note length or rest rest length. You may combine the note importance according to the position of a note. For example,
A value obtained by calculating from both values such as the sum or product of the value determined by the note length of the note and the value determined by the position of the note can be used as the note importance. Furthermore, when the note in the clue note sequence is given note reliability, the note reliability of the note in the clue note sequence may be set as the note importance. Not only the reliability of the note, but also the sum or product of the note reliability of the immediately preceding note and the note importance of the two notes may be increased as the note reliability of the two notes increases.

The similarity calculation means 6 is the note correspondence calculation means 3
The note correspondences are received from the note matching degree calculation means 4
From the note importance degree calculation means 5 receives the note importance degree. Next, based on the correspondence between the notes, the degree of similarity between the note strings is obtained from the values of the note coincidence and the note importance. As a method of calculating the similarity between note strings, for example, the product of the note coincidence and the note importance is calculated for each correspondence between notes or rests based on the correspondence relationship, and the sum of these values is referred to as the similarity between note strings. There is something to do.

The output means 7 receives the similarity between note sequences from the similarity calculation means 6 and outputs it to another device such as a display device / printer device / storage device.

The calculation formulas and calculation parameters used for the above calculation can be set independently for each means. For these settings, the user or the designer of the music retrieval system shall set them in each device in advance prior to the operation.

1, FIG. 3, FIG. 4 and FIG. 5 (A), in the present embodiment, a method of comparing notes in order from the beginning of a note string is used as a note correspondence method. As the note importance, the note length of the corresponding note in the clue note sequence and the note length of the immediately preceding note are used as the note importance. The operation will be described using a specific example, taking as an example the case where the sum of the products of the note coincidence and the note importance of the corresponding notes is used as the similarity between the note sequences.

The cue note string 1 of FIG. 4 is input to the cue note string input means 1 from the outside (FIG. 3, step A).
1). Next, the candidate note string input means 2 is externally input with the candidate note string 1 of FIG. 4 (FIG. 3, step A2). Next, the note correspondence calculation means 3 obtains note correspondences in which the notes are compared in order from the beginning of each of the clue note string 1 and the candidate note string 1 (FIG. 3, step A3). Next, the note coincidence calculation means 4 compares the pitch difference with the immediately preceding note between the clue note sequence 1 and the candidate note sequence 1, and obtains the note coincidence degree based on each note correspondence (FIG. 3, step A4). . Next, the note importance calculation means 5 sets the product of the note length of the corresponding note in the clue note string 1 and the note length of the note immediately preceding it as the note importance (FIG. 3, step A5). Next, the similarity calculation means 6 obtains the product of the note coincidence and the note importance from the correspondence between the notes, the note coincidence, and the note importance, and calculates the sum of the products, 1.75, between the note strings. (Fig. 3,
Step A6). Next, the output means 7 outputs the similarity between note strings to the outside (FIG. 3, step A7).

In the above description of the operation, the note string input is performed in the order of step A1 and step A2, but the note string input may be performed in the order of step A2 and step A1. The note sequence input of A1 and step A2 may be performed simultaneously. Furthermore, in the above description of the operation, the calculation is performed in the order of step A4 and step A5, but the calculation may be performed in the order of step A5 and step A4, and the calculation of step A4 and step A5. May be performed simultaneously.

Further, as shown in FIG. 5B, the clue note string 1 is input to the clue note string input means 1 and the candidate note string 2
Is given to the candidate note string input means 2, the similarity between note strings can be calculated as 3.5 by the same calculation as described above. By using this embodiment in this way,
Comparing candidate note sequence 1 and candidate note sequence 2 results in candidate note sequence 2
It is possible to obtain the similarity between the note strings, that is more similar to the clue note string 1.

Next, referring to FIGS. 1, 3, 4 and 6A, a method of comparing notes in the present embodiment, which compares notes in order from the beginning of a note string, is used as a note correspondence method. 1 if the pitch difference from the immediately preceding sound matches, 0 if they do not match
2 is used as the note importance degree, and the note importance degree shown in FIG. 2A is used according to the note position in the clue note string, and as the note-to-note string similarity degree, the corresponding notes are matched. Taking the case of using the sum of products of degrees and note importance as an example, the operation will be described using a specific example.

The cue note string 1 of FIG. 4 is input to the cue note string input means 1 from the outside (FIG. 3, step A).
1). Next, the candidate note string input means 2 is externally input with the candidate note string 1 of FIG. 4 (FIG. 3, step A2). Next, the note correspondence calculation means 3 obtains note correspondences in which the notes are compared in order from the beginning of each of the clue note string 1 and the candidate note string 1 (FIG. 3, step A3). Next, the note coincidence calculation means 4 compares the pitch difference with the immediately preceding note between the clue note sequence 1 and the candidate note sequence 1, and obtains the note coincidence degree based on each note correspondence (FIG. 3, step A4). . Next, the note importance calculation means 5 determines the note importance according to the position of the note in the clue note sequence 1 between the corresponding notes (FIG. 3, step A5). Next, the similarity calculation means 6
Is the sum of the products obtained by finding the product of the note coincidence and the note importance from the note correspondence, the note coincidence and the note importance.
Is the similarity between note strings (FIG. 3, step A6). Next, the output means 7 outputs the similarity between note strings to the outside (FIG. 3, step A7).

In the above description of the operation, the note string input is performed in the order of step A1 and step A2, but the note string input may be performed in the order of step A2 and step A1. The note sequence input of A1 and step A2 may be performed simultaneously. Furthermore, in the above description of the operation, the calculation is performed in the order of step A4 and step A5, but the calculation may be performed in the order of step A5 and step A4, and the calculation of step A4 and step A5. May be performed simultaneously.

Further, as shown in FIG. 6B, the clue note string 1 is input to the clue note string input means 1 and the candidate note string 2 is added.
Is given to the candidate note string input means 2, the similarity between note strings can be obtained as 2.5 by the same calculation as described above. By using this embodiment in this way,
Comparing candidate note sequence 1 and candidate note sequence 2 results in candidate note sequence 2
It is possible to obtain the similarity between the note strings, that is more similar to the clue note string 1.

Next, referring to FIGS. 1, 3, 4 and 7A, a method of comparing notes in order from the beginning of a note string is used as a note correspondence method in this embodiment, and the degree of note coincidence is used. Is set to 1 if the pitch difference from the immediately preceding note matches, and 0 if not to match. The note importance is the product of the corresponding note of the clue note sequence and the note reliability of the note immediately before the note. As an example of the similarity between the note strings, the case where the sum of the products of the note coincidence and the note importance of the corresponding notes is used is taken as an example, and the operation will be described using a concrete example.

The cue note string 1 of FIG. 4 is input to the cue note string input means 1 from the outside (FIG. 3, step A).
1). Next, the candidate note string input means 2 is externally input with the candidate note string 1 of FIG. 4 (FIG. 3, step A2). Next, the note correspondence calculation means 3 obtains note correspondences in which the notes are compared in order from the beginning of each of the clue note string 1 and the candidate note string 1 (FIG. 3, step A3). Next, the note coincidence calculation means 4 compares the pitch difference with the immediately preceding note between the clue note sequence 1 and the candidate note sequence 1, and obtains the note coincidence degree based on each note correspondence (FIG. 3, step A4). . Next, the note importance calculation means 5 determines the note importance between the corresponding notes based on the note reliability of the clue note string (FIG. 3, step A5). Next, the similarity calculation means 6
The product of the note coincidence and the note importance is obtained from the note correspondence, the note coincidence, and the note importance, and the sum of the products is 3.2.
5 is set as the similarity between note strings (FIG. 3, step A6). Next, the output means 7 outputs the similarity between note strings to the outside (FIG. 3, step A7).

In the above description of the operation, the note string input is performed in the order of step A1 and step A2, but the note string input may be performed in the order of step A2 and step A1. The note sequence input of A1 and step A2 may be performed simultaneously. Furthermore, in the above description of the operation, the calculation is performed in the order of step A4 and step A5, but the calculation may be performed in the order of step A5 and step A4, and the calculation of step A4 and step A5. May be performed simultaneously.

As shown in FIG. 7B, the clue note string 1 is input to the clue note string input means 1 and the candidate note string 2 is input.
Is given to the candidate note string input means 2, the similarity between note strings can be obtained as 7 by the same calculation as described above. By using the present embodiment as described above, it is possible to obtain the inter-note-note similarity degree that the candidate note string 2 is more similar to the clue note string 1 when the candidate note string 1 and the candidate note string 2 are compared. .

As described above, when comparing the cue note sequence and the candidate note sequence, the note importance is calculated from the note length, note position, note reliability, etc. of the cue note sequence, and the note sequence similarity is calculated. By calculating the value of, it becomes possible to calculate the degree of similarity more appropriately than in the past. As an example of how to calculate the note importance, we have shown calculation examples of note length, note position, and note reliability, but it is of course possible to combine the above values in note importance calculation. . For example, the product of the note importances obtained by the above-described method or the linear sum may be used.

FIG. 8 is a block diagram showing another embodiment of the present invention. This embodiment is in addition to the embodiment shown in FIG.
There is provided a note string deformation deduction calculation means 11 for obtaining a note string deformation deduction point due to a note string deformation. Further, the note correspondence calculation means 3 transforms the note sequence when obtaining the correspondence between the note sequences, and obtains a plurality of correspondences. Furthermore, the similarity calculation means 6
However, the degree of similarity is calculated not only by the degree of note coincidence and the degree of similarity correction, but also by including the note string deformation deduction points.

The note correspondence calculation means 3 receives the clue note string from the clue note string input means 1 and the candidate note string input means 2
Each of the candidate note strings is received from each of the received note strings, and a plurality of pairs of received clue note strings and candidate note string variants are obtained. The modification of the musical note sequence can be realized by repeating an operation of inserting a virtual musical note that is not a comparison target at any musical note position in any of the received musical note sequences an arbitrary number of times. Virtual notes inserted for the purpose of changing note correspondence are not used for note comparison between note strings. Therefore, it is not necessary to have the attribute values such as the pitch value and the pitch length value. The note correspondence is obtained between notes having attribute values included in the clue note sequence and the candidate note sequence. As an example of how to find the note correspondence, when i is a natural number, and when both i-th notes from the beginning of the note sequence have notes with attribute values, the correspondence between these notes, If both the i-th note or rest in the order from the beginning are between notes with attribute values or between rests with attribute values, those that correspond to these notes or rests, in order from the end of the note string When both the i-th note and the rest are between notes having an attribute value or between rests having an attribute value, it is possible to use a correspondence between these notes or rests.

The note string deformation deduction calculation means 11 receives a clue note string from the clue note string input means 1, a candidate note string from the candidate note string input means 2, and a note correspondence relationship from the note correspondence calculation means 3. From the correspondence between the received notes, a note string deformation deduction point, which is a deduction point according to the degree of deformation of the note string, is obtained. The note sequence modification deduction is, for example, the value obtained by multiplying the number of note insertions performed for modification operation by a negative coefficient, or the value obtained by multiplying the note length of unmatched notes generated by note insertion by a negative coefficient. As described above, it is possible to use a value such that the absolute value of the deduction points increases according to the degree of deformation.

The similarity calculation means 5 is the note correspondence calculation means 3
From the correspondence of the notes from the note matching degree calculation means 3
From the note importance degree calculation means 4 and from the note string deformation deduction calculation means 11 from the note string deformation deduction point. Based on the correspondence of each note, the correspondence similarity is obtained from the note coincidence, the note importance, and the note string deformation deduction point, and the inter-note string similarity is obtained from the correspondence similarities obtained for all the correspondences obtained by the note correspondence calculation means 3. Ask.
As a method of obtaining the correspondence similarity, for example, the product of the note coincidence and the note importance is obtained for each correspondence between notes based on the correspondence, and the sum of this value and the sum of the note string deformation deduction points is taken as the correspondence similarity. There is something to do. Examples of the similarity between note sequences include the average value, the median value, the maximum likelihood value, and the maximum value of the corresponding similarity degrees.

Naturally, the calculation formulas and calculation parameters used for the above calculation can be set independently for each means. For these settings, the user or the designer of the music retrieval system shall set them in each device in advance prior to the operation.

Next, with reference to FIGS. 8, 9 and 10 (A), in the present embodiment, a note transformed from the original note sequence by inserting a note at each note position in correspondence with the note. A plurality of columns are obtained, and the note correspondence is used in which the notes are compared in order from the beginning between the modified note strings, and the note matching degree is set to 1 if the pitch difference with the immediately preceding note matches, and 0 if not matched. As the note importance, the product of the note length of the corresponding note in the clue note sequence and the note length of the immediately preceding note is used as the note importance, and -1
As a point deduction, the case of using the maximum value of the correspondence similarity obtained by the sum of the products of the note coincidence and the note importance of the corresponding notes as an example, and using a concrete example The operation will be described.

The cue note string 1 is input to the cue note string input means 1 (FIG. 9, step B1). Next, the candidate note string 1 is input to the candidate note string input means 2 (FIG. 9, step B2). Next, the note correspondence calculation means 3 inserts a note at each note position as a note correspondence, obtains a plurality of modified note strings from the original note string, and sequentially notes from the beginning among the modified note strings. Then, the correspondence of the notes is compared (step B3 in FIG. 9). Next, the note coincidence calculation means 4 compares the pitch difference with the immediately preceding note between the note sequences, and obtains the note coincidence based on each note correspondence (FIG. 9, step B4). Next, the note importance calculation means 5 sets the product of the note length of the corresponding note in the clue note sequence 1 and the note length of the immediately preceding note between the corresponding notes as the note importance (FIG. 9, step). B5). Next, the musical score sequence deformation deduction calculation means 1
1 obtains a note string deformation deduction point according to the note string deformation (FIG. 9, step B6). Next, the similarity calculation means 6
The correspondence similarity based on the correspondence of the notes is obtained from the correspondence of the notes, the degree of coincidence of the notes, and the degree of importance of the notes (FIG. 9, step B).
7). Furthermore, if there is another correspondence relationship calculated by the note correspondence calculation means 3, the correspondence similarity is calculated for the other note correspondence relationships (FIG. 9, step B8). Then, from the correspondence similarities based on all the note correspondences, 4 is obtained as the inter-note string similarity (FIG. 9, step B9). Finally,
The output means 7 outputs the similarity between note strings to the outside (FIG. 9, FIG.
Step B10).

In the above description of the operation, note strings are input in the order of step B1 and step B2, but note string input may be carried out in the order of step B2 and step B1. And the note string input in step B2 may be performed simultaneously. Furthermore, in the above description of the operation, the calculation is performed in the order of step B4, step B5, and step B6. Calculations may be performed in order from any step, and
Calculations may be performed at the same time. Further, when a plurality of correspondences are calculated by the note correspondence calculating means 3, it is possible to obtain the correspondence relationship and the correspondence similarity and the similarity between note strings while deforming the note string. In such cases,
The degree of similarity between musical note sequences can be obtained using dynamic programming. The dynamic programming method is described in detail, for example, in the document "Basics of Speech Information Processing", pages 187 to 190 (published by Ohmsha).

Further, in FIG. 10B, the clue 1
When the note sequence and the candidate note sequence 3 are given to the input means 1, the inter-note sequence similarity can be obtained as 7 by the same calculation as described above. By using the present embodiment as described above, it is possible to obtain the inter-note-note similarity degree that the candidate note string 3 is more similar to the clue note string 1 when the candidate note string 1 and the candidate note string 3 are compared. .

[0043]

As described above, according to the present invention, it is possible to use the degree of importance as a clue for the notes constituting the note sequence which serves as a clue for retrieval. Therefore, by using the present invention, it is possible to calculate an appropriate degree of similarity according to the importance of a note, as compared with a conventional apparatus in which all the notes in a cue note string are treated with the same importance, and as a result, use It is possible to search for a music effectively using a note sequence reproduced based on a person's memory as a clue.

[Brief description of drawings]

FIG. 1 is a block diagram showing a first embodiment of the present invention.

FIG. 2 is an explanatory diagram for explaining an example of an operation of the note importance calculation means 5.

FIG. 3 is a flow chart for explaining the first embodiment of the present invention.

FIG. 4 is a diagram showing an example of a note sequence input to a note sequence similarity calculation device.

FIG. 5 is an explanatory diagram showing an example of the operation of the first exemplary embodiment of the present invention.

FIG. 6 is an explanatory diagram showing an example of the operation of the first exemplary embodiment of the present invention.

FIG. 7 is an explanatory diagram showing an example of the operation of the first exemplary embodiment of the present invention.

FIG. 8 is a block diagram showing a second embodiment of the present invention.

FIG. 9 is a flowchart for explaining a second embodiment of the present invention.

FIG. 10 is an explanatory diagram showing an operation of the second exemplary embodiment of the present invention.

FIG. 11 is a block diagram showing a conventional example.

FIG. 12 is an explanatory diagram showing an operation of a conventional example.

[Explanation of symbols]

 1 clue note string input means 2 candidate note string input means 3 note correspondence calculating means 4 note coincidence calculating means 5 note importance calculating means 6 similarity calculating means 7 output means 11 note string deformation deduction calculating means

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Internal reference number FI technical display location G11B 27/28 F 8224-5D

Claims (1)

[Claims] 1. A cue note sequence inputting means for inputting the cue note sequence and a candidate for inputting the candidate note sequence in a note sequence similarity calculator for comparing and collating a cue note sequence and a candidate note sequence. Note string input means, note correspondence calculation means for obtaining correspondence of notes to be compared between the clue note string and the candidate note string, and note matching in note units between the clue note string and the candidate note string based on the correspondence And a note importance degree calculating means for obtaining the note importance degree of each note of the clue note sequence based on the correspondence, and a note sequence from the note agreement degree and the note importance degree based on the correspondence An inter-note-note-similarity calculating device comprising a similarity calculating means for calculating an inter-note similarity and an output means for outputting the inter-note-string similarity