JP2002278547A - Music piece retrieval method, music piece retrieval data registration method, music piece retrieval device and music piece retrieval data registration device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method and a device for making a desired music piece retrievable by a user, on the basis of a subjective impression of the music piece. SOLUTION: The subjective requirements of the user for the music piece whose retrieval is desired are inputted, digitized and outputted, a predicted impression value for which the impression of the music piece of a retrieval object is calculated from the output and a music piece data base which store the acoustic signals of a plurality of the music pieces and impression values for which the impressions of the music pieces are quantified is retrieved with the predicted impression value as a key. Thus, on the basis of the subjective image for the music piece of the user, the desired music piece are mode retrievable.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a music search method, a music search device, and music search data for searching for and presenting a music desired by a user based on bibliographic information on the music and a subjective impression of the music. The present invention relates to a registration method, a music search data registration method, a music search program, and a music search data registration program.

[0002]

2. Description of the Related Art When searching for a specific song from a song database storing a large amount of music, such as a hard disk or a CD player with a serial connection function, generally, the title of the song, the name of a singer, the name of a lyricist / composer, etc. In many cases, a search is performed using the keyword as a keyword.

A conventional general music search apparatus will be described with reference to FIG. In FIG. 8, reference numeral 81 denotes music selection condition input means for inputting music selection conditions, such as a music title to be searched. Reference numeral 82 denotes a music database, in which bibliographic information such as a music title, a singer name, a lyric / composer name, etc. of each music is recorded together with an audio signal of the music, and can be referred to. Reference numeral 83 denotes a bibliographic information search means, which refers to a music database using conditions such as a music title input by the music selection condition input means 81 as a search key, and outputs music having bibliographic information that matches or is similar to the search key.
Numeral 84 denotes a reproduction music selecting means for selecting a music desired by the user to reproduce from the search result of the bibliographic information searching means 83. Numeral 85 is a music reproducing means for reproducing the music selected by the reproducing music selecting means 84 so as to be audible.

[0004] The operation of a conventional music search device will be described with an example. Now, it is assumed that the user wants to search for and listen to music related to “Christmas”. In this case, the user generally inputs "Christmas" in the item of the song name using the song selection condition input means 81, and searches for songs whose title includes the character string "Christmas". The bibliographic information search means 83 refers to the music database 82, searches for music in which the character string "Christmas" is included in the bibliographic information using the music title "Christmas" as a search key, and outputs the result. As a result of the search, assuming that three songs “Christmas Carol”, “Christmas Eve”, and “Merry Christmas on the battlefield” are output, the reproduction song selection unit 84 provides other bibliographic information (singer name, songwriting / composition) of these three songs The user can select a tune to be reproduced with reference to the name of the user), and if there is a desired tune, the tune is reproduced and listened to by the tune reproduction means 85.

[0005]

However, the above-mentioned bibliographic information such as the song title and singer name of the song is information from an objective and external viewpoint of the song, and the song itself is provided to the listener. It is difficult to guess the subjective impression. For this reason, there is a problem that it is difficult to search for music based on a subjective impression such as "I want to search for music with a lively feeling" using a conventional music search device. On the other hand, in order to realize a music search based on the subjective impression, a subjective impression of the music is indexed and registered in the music database 82 in advance, and a desired music is searched based on the index information. A method is conceivable. However, in this case, the work of registering the index information requires a person (mainly the creator or user of the apparatus) to listen to each song and judge its impression, and when there are many songs to be registered, the work is required. There was a problem that the burden also increased.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a method and an apparatus for enabling a user to search for a desired music based on a subjective impression of the music.

[0007]

In order to achieve the above object, a music search method according to claim 1 of the present invention inputs a user's subjective requirement for a music desired to be searched and converts it into a numerical value. A search requirement input step to be output, a predicted impression value calculation step of calculating a predicted impression value by quantifying the impression of the search target song from the output of the search requirement input step, and a sound signal of a plurality of music pieces and a sound signal of the music piece. A song database storing impression values obtained by quantifying impressions, and an impression value search step of searching the song database using the predicted impression value as a key is provided.

According to a second aspect of the present invention, in the music search method according to the first aspect, an audio signal inputting step of inputting an audio signal of the music, and a predetermined characteristic amount is extracted from the audio signal. Volume extraction step, and an impression value calculation step of calculating an impression value obtained by quantifying the impression of the music from the feature amount, wherein the music database includes:
The sound signal and the impression value are stored.

According to a third aspect of the present invention, in the music search method of the first or second aspect, the search requirement input step includes a music impression value calculation step of calculating an impression value of the input music. An audio signal inputting step of inputting an audio signal of a music to be searched; a feature extracting step of extracting a predetermined feature from the audio signal; An impression value calculating step of calculating an impression value, wherein the music database is searched using the impression value calculated in the music impression value calculating step as a key.

According to a fourth aspect of the present invention, in the music search method according to the second aspect, the music database includes:
The feature amount extracted in the feature extraction step is stored, the search requirement input step inputs a search requirement related to the feature amount of the search target song, and the impression value search step includes a search amount related to the feature amount of the search target song. The music database is searched using a requirement as a search key.

According to a fifth aspect of the present invention, in the music search method according to the second aspect, the impression value calculating step is performed by multivariate analysis or the like from feature amounts of a plurality of music pieces and impression values of the music pieces. And calculating an impression value of the music based on the coefficient.

According to a sixth aspect of the present invention, in the music search method according to any one of the second to fifth aspects, the feature amount extracting step includes determining a sound period and a sound intensity included in an audio signal of the music. The feature amount is calculated as a feature amount.

According to a seventh aspect of the present invention, in the music search method according to any one of the second to fifth aspects, the feature amount extracting step includes the step of expressing the quantity representing the degree of change in the spectrum of the audio signal of the music feature amount. Is calculated as

According to an eighth aspect of the present invention, in the music search method according to any one of the second to fifth aspects, the feature amount extracting step includes, as a feature amount, a quantity expressing a sounding frequency of an audio signal of the music piece. It is characterized by calculating.

According to a ninth aspect of the present invention, in the music search method according to any one of the second to fifth aspects, the feature amount extracting step includes calculating a quantity representing a degree of sound periodicity of an audio signal of the music. The feature amount is calculated as a feature amount.

The invention according to claim 10 is the same as the invention according to claim 2.
5. The music retrieval method according to any one of items 1 to 5, wherein the feature amount extraction step calculates a quantity representing a beat cycle of the music as a feature amount.

The invention according to claim 11 is the same as the invention according to claim 2.
6. The music search method according to any one of claims 1 to 5, wherein the feature amount extracting step calculates a quantity expressing a beat cycle ratio of the music as a feature amount.

The invention according to claim 12 is the invention according to claim 2.
6. The music search method according to any one of claims 1 to 5, wherein the feature amount extracting step calculates a quantity representing a beat intensity of the music as a feature amount.

The invention according to claim 13 is the invention according to claim 2.
5. The music retrieval method according to any one of items 1 to 5, wherein the feature amount extracting step calculates a quantity expressing a ratio of beat intensities of the music as a feature amount.

The invention according to claim 14 is the same as the invention according to claim 2.
6. The music search method according to any one of claims 1 to 5, wherein the feature amount extracting step divides an audio signal of the music into a plurality of bands, detects a rising component of each band, and uses the rising component of each band to detect the feature. The amount is calculated.

The invention according to claim 15 is the first invention.
15. In the music search method according to any one of to 14, the predictive impression value calculating step calculates the predictive impression value based on a predetermined quantity according to the subjective requirements input in the search requirement input step. It is characterized by.

The invention according to claim 16 is the first invention.
15. The music search method according to any one of claims 15 to 15, wherein the impression value search step calculates a Euclidean distance between the predicted impression value and an impression value of each music stored in the music database, Are output as candidate songs in order from the song with the smallest number.

[0023] The invention according to claim 17 is based on claim 1.
15. The music search method according to any one of claims 15 to 15, wherein the impression value search step calculates a Euclidean distance between the predicted impression value and an impression value of each music stored in the music database, Are output as candidate tunes in order from a tune having a smaller Euclidean distance for tunes shorter than a predetermined distance.

The invention according to claim 18 is the first invention.
18. The music search method according to 6 or 17, wherein the impression value search step calculates a similarity between the predicted impression value and the impression value of the candidate music based on the Euclidean distance, and outputs the similarity together with the candidate music. It is characterized by the following.

The invention according to claim 19 is the first invention.
19. The music search method according to any one of items 1 to 18, wherein information on the candidate music output from the impression value search step is extracted from the music database and presented to the user, and a music is selected from the candidate music. And a music reproduction step of extracting an audio signal of the music selected in the music selection step from the database and reproducing a part or all of the music so that a user can listen to the music.

The invention according to claim 20 is the first invention.
9. The music search method according to 9, wherein the music selection step includes:
A feature is that music is selected according to a predetermined music selection method.

Further, the invention according to claim 21 is based on claim 1
20. The music search method according to any one of claims 1 to 20, wherein the search requirement inputting step inputs requirements relating to bibliographic information relating to music, such as a music title, the music database stores bibliographic information relating to music, and the impression value searching step. Is characterized in that the music database is searched using the bibliographic information as a key.

Further, the invention according to claim 22 is based on claim 2
1. The music search method according to item 1, further comprising a bibliographic information input step of inputting bibliographic information on music such as a music title, wherein the music database stores the bibliographic information.

[0029] The invention according to claim 23 is based on claim 2.
23. The music search method according to 1 or 22, wherein the impression value search step uses the impression value of the music corresponding to the bibliographic information input in the search requirement input step as a search key as a predicted impression value. .

The invention according to claim 24 is the same as the claim 2.
24. The music search method according to any one of 1 to 23, wherein the search requirement input step inputs a search method of a requirement relating to bibliographic information such as a song title and a requirement relating to subjectivity such as an image of a song, and the impression value retrieval step is performed in accordance with the search method. The search for bibliographic information and the search for a predicted impression value are performed in combination.

In the music search data registration method according to a twenty-fifth aspect of the present invention, an audio signal inputting step of inputting an audio signal of a music and an audio signal outputted by the audio signal inputting step are characterized by a predetermined characteristic. A feature value extracting step of extracting a quantity; and an impression value calculating step of calculating an impression value obtained by quantifying an impression of the music from the feature quantity.

The invention according to claim 26 is based on claim 2
5. The method of registering music search data according to claim 5, wherein the feature amount extracting step calculates, as a feature amount, a quantity related to a sounding period and a sounding intensity included in an audio signal of the music piece.

[0033] The invention according to claim 27 is based on claim 2.
5. The method of registering music search data according to claim 5, wherein the feature amount extracting step calculates, as the feature amount, a quantity expressing the degree of change in the spectrum of the audio signal of the music piece.

The invention according to claim 28 is the second invention.
5. The method for registering music search data according to 5, wherein the feature amount extracting step calculates a quantity expressing a sounding frequency of an acoustic signal of the music as a feature amount.

Further, the invention described in claim 29 is based on claim 2
5. The music search data registration method according to 5, wherein the feature amount extracting step calculates, as a feature amount, a quantity expressing a degree of sound periodicity of an audio signal of the music.

The invention according to claim 30 is the same as the claim 2.
5. The music search data registration method according to 5, wherein the feature amount extracting step calculates a quantity representing a beat cycle of the music as a feature amount.

The invention according to claim 31 is based on claim 2
5. The music search data registration method according to 5, wherein the feature amount extracting step calculates a quantity representing a beat cycle ratio of the music as a feature amount.

The invention described in claim 32 is the second invention.
5. The music search data registration method according to 5, wherein the feature amount extracting step calculates a quantity expressing the beat intensity of the music as a feature amount.

The invention according to claim 33 is the second invention.
5. The music search data registration method according to 5, wherein the feature amount extracting step calculates a quantity representing a ratio of beat intensities of the music as a feature amount.

The invention described in claim 34 is the second invention.
5. In the music search data registration method according to 5, the feature amount extracting step divides an audio signal of the music into a plurality of bands, detects a rising component of each band, and uses the rising component of each band to calculate a feature amount. Is calculated.

Further, in order to achieve the above object, a music search device according to claim 35 of the present invention provides a search requirement in which a user's subjective requirement for a music desired to be searched is input, quantified and output. Input means, a predicted impression value calculating means for calculating a predicted impression value obtained by quantifying the impression of the music to be searched from the output of the search requirement input means, and quantifying the sound signals of the plurality of music pieces and the impressions of the music pieces A music database that stores the impression values obtained by the above-described method, and an impression value search unit that searches the music database using the predicted impression value as a key.

The invention according to claim 36 is based on claim 3
5. The music search device according to claim 5, wherein: an audio signal input unit for inputting an audio signal of the music; a characteristic amount extraction unit for extracting a predetermined characteristic amount from the audio signal; and an impression of the music based on the characteristic amount. Impression value calculation means for calculating a quantified impression value, wherein the music database stores the sound signal and the impression value.

The invention according to claim 37 is the third invention.
37. The music search device according to 5 or 36, wherein the search requirement input means has a music impression value calculation means for calculating an impression value of the input music, and the music impression value calculation means includes An audio signal input unit for inputting an audio signal, a characteristic amount extraction unit for extracting a predetermined characteristic amount from the audio signal, and an impression value calculation unit for calculating an impression value of a music piece from the characteristic amount; The music database is searched using the impression value calculated by the music impression value calculation means as a key.

The invention according to claim 38 is the third invention.
7. The music search device according to claim 6, wherein the music database stores the feature quantity extracted by the feature extraction means, the search requirement input means inputs a search requirement related to a feature quantity of a search target music, and the impression value A search unit searches the music database using a requirement relating to a feature amount of the music to be searched as a search key.

The invention according to claim 39 is directed to claim 3
7. The music search device according to claim 7, wherein the impression value calculation means calculates an impression value of the music based on a feature amount of the plurality of music and a coefficient calculated by a multivariate analysis from the impression value of the music. It is characterized by.

The invention according to claim 40 is the third invention.
40. The music search device according to any one of items 6 to 39, wherein the feature amount extracting means calculates, as a feature amount, a quantity related to a sounding period and a sounding intensity included in an audio signal of the music piece.

The invention according to claim 41 is the third invention.
40. The music search device according to any one of items 6 to 39, wherein the feature amount extracting means calculates a quantity expressing a degree of change in the spectrum of an audio signal of the music as a feature amount.

The invention according to claim 42 is the third invention.
40. The music search device according to any one of items 6 to 39, wherein the feature amount extracting means calculates a quantity expressing a sounding frequency of an audio signal of the music as a feature amount.

The invention according to claim 43 is the third invention.
40. The music search device according to any one of items 6 to 39, wherein the feature amount extracting means calculates a quantity expressing a degree of a sound periodicity of an audio signal of the music as a feature amount.

The invention according to claim 44 is the third invention.
40. The music search device according to any one of items 6 to 39, wherein the feature amount extracting means calculates a quantity representing a beat cycle of the music as a feature amount.

The invention according to claim 45 is the third invention.
40. The music search device according to any one of items 6 to 39, wherein the feature amount extracting means calculates a quantity expressing a beat cycle ratio of the music as a feature amount.

The invention according to claim 46 is directed to claim 3
40. The music search device according to any one of items 6 to 39, wherein the feature amount extracting means calculates a quantity expressing a beat intensity of the music as a feature amount.

The invention described in claim 47 is the third invention.
40. The music search device according to any one of items 6 to 39, wherein the feature amount extracting means calculates a quantity expressing a ratio of beat intensities of the music as a feature amount.

The invention according to claim 48 is the third invention.
40. The music search device according to any one of 6 to 39, wherein the feature amount extracting unit divides an audio signal of the music into a plurality of bands, detects a rising component of each band, and uses the rising component of each band. The feature amount is calculated.

The invention according to claim 49 is the third invention.
49. The music search device according to any one of items 5 to 48, wherein the predicted impression value calculation means calculates a predicted impression value based on a predetermined quantity according to the subjective requirement input by the search requirement input means. It is characterized by the following.

The invention described in claim 50 is the third invention.
49. The music search apparatus according to any one of items 5 to 48, wherein the impression value search means calculates a Euclidean distance between the predicted impression value and an impression value of each music stored in the music database, and It is characterized in that music pieces with smaller distances are output as candidate music pieces in order.

The invention described in claim 51 is the third invention.
50. The music search device according to any one of items 5 to 49, wherein the impression value search means calculates a Euclidean distance between the predicted impression value and an impression value of each music stored in the music database, and For songs whose distance is equal to or less than a predetermined distance, the songs are output as candidate songs in order from the song having the smallest Euclidean distance.

The invention according to claim 52 is the same as the invention according to claim 5.
53. In the music search device according to 0 or 51, the impression value search means calculates a similarity between the predicted impression value and the impression value of the candidate music based on the Euclidean distance, and outputs the similarity together with the candidate music. It is characterized by the following.

The invention according to claim 53 is the third invention.
53. The music search apparatus according to any one of items 5 to 52, wherein information on the candidate music output from the impression value search means is extracted from the music database and presented to a user, and music is selected from the candidate music. And music reproducing means for extracting an audio signal of the music selected by the music selecting means from the database and reproducing a part or all of the music so that a user can listen to the music.

Further, the invention according to claim 54 is directed to claim 5
3. The music search device according to 3, wherein the music selection means selects a music according to a predetermined music selection device.

The invention according to claim 55 is the third invention.
55. The music search device according to any one of items 5 to 54, wherein the search requirement input means inputs requirements relating to bibliographic information regarding music such as a music title, the music database stores bibliographic information regarding music, and the impression value search. Means for searching the music database using the bibliographic information as a key.

The invention according to claim 56 is directed to claim 5
5. The music search device according to 5, wherein the music database has bibliographic information input means for inputting bibliographic information on music such as a music title, and the music database stores the bibliographic information.

The invention according to claim 57 is based on claim 5
56. The music search device according to item 5 or 56, wherein the impression value search means uses the impression value of the music corresponding to the bibliographic information input by the search requirement input means as a search key as a predicted impression value. .

The invention according to claim 58 is based on claim 5
58. The music search device according to any one of items 5 to 57, wherein the search requirement input means inputs a search device for requirements relating to bibliographic information such as a song title and a requirement relating to subjectivity such as an image of a song. It is characterized in that a search for bibliographic information and a search for a predicted impression value are performed in combination depending on the device.

A music registration data registration device according to claim 59 of the present invention is characterized in that a predetermined characteristic amount is determined from an audio signal input means for inputting an audio signal of a music and an audio signal output from the audio signal input means. And an impression value calculating means for calculating an impression value obtained by quantifying the impression of the music piece from the feature amount.

The invention according to claim 60 is the same as that in claim 5.
10. The music search data registration device according to Item 9, wherein the feature amount extracting means calculates, as a feature amount, a quantity related to a sounding period and a sounding intensity included in an audio signal of the music piece.

Further, the invention according to claim 61 is based on claim 5
9. The music search data registration device according to Item 9, wherein the feature amount extracting means calculates a quantity representing a degree of change in the spectrum of an audio signal of the music as a feature amount.

The invention according to claim 62 is based on claim 5
9. The music search data registration device according to Item 9, wherein the feature amount extracting means calculates a quantity expressing the sounding frequency of an audio signal of the music as a feature amount.

The invention according to claim 63 is the same as that in claim 5
10. The music search data registration device according to Item 9, wherein the feature amount extracting means calculates a quantity expressing a degree of sound periodicity of an acoustic signal of the music as a feature amount.

The invention according to claim 64 is based on claim 5
9. The music search data registration device according to Item 9, wherein the feature amount extracting means calculates a quantity representing a beat cycle of the music as a feature amount.

The invention according to claim 65 is the same as that in claim 5
10. The music search data registration device according to claim 9, wherein the feature amount extracting means calculates a quantity expressing a beat cycle ratio of the music as a feature amount.

Further, the invention according to claim 66 is based on claim 5
9. The music search data registration device according to Item 9, wherein the feature amount extracting means calculates a quantity expressing the beat intensity of the music as a feature amount.

The invention according to claim 67 is the same as that in claim 5
10. The music search data registration device according to Item 9, wherein the feature amount extracting means calculates a quantity expressing a ratio of beat intensities of the music as a feature amount.

The invention described in claim 68 is the same as that in claim 5
9. In the music search data registration device according to Item 9, the feature amount extraction means divides an audio signal of the music into a plurality of bands, detects a rising component of each band, and uses the rising component of each band to calculate a feature amount. Is calculated.

Further, in order to achieve the above object, a music search program according to claim 69 of the present invention provides a computer which inputs a user's subjective requirement for a music desired to be searched, converts it into a numerical value, and outputs it. Search requirement input means, predicted impression value calculation means for calculating a predicted impression value obtained by quantifying the impression of the search target music from the output of the search requirement input means, sound signals of a plurality of music pieces, and impressions of the music pieces And a function of functioning as a music database storing impression values obtained by quantifying the above, and an impression value search means for searching the music database using the predicted impression value as a key.

The invention described in claim 70 is the same as that in claim 6
9. The music search program according to claim 9, wherein: an audio signal input means for inputting an audio signal of the music; a characteristic amount extraction means for extracting a predetermined characteristic amount from the audio signal; Impression value calculation means for calculating a quantified impression value, wherein the music database is made to function so as to store the acoustic signal and the impression value.

The invention described in claim 71 is the same as that in claim 6
70. In the music search program according to 9 or 70, the search requirement input means has a music impression value calculation means for calculating an impression value of the input music, and the music impression value calculation means has An audio signal input unit for inputting an audio signal, a characteristic amount extraction unit for extracting a predetermined characteristic amount from the audio signal, and an impression value calculation unit for calculating an impression value of a music piece from the characteristic amount; A computer function is made to search the music database using the impression value calculated by the music impression value calculation means as a key.

Further, the invention according to claim 72 is based on claim 7
0, the song database stores the feature amount extracted by the feature extracting unit, the search requirement input unit inputs a search requirement related to a feature amount of a search target song, and the impression value The search means causes a computer to function so as to search the music database using a requirement relating to the feature amount of the music to be searched as a search key.

Further, the invention according to claim 73 is the same as that in claim 7
In the music search program according to the first aspect, the impression value calculation means may calculate an impression value of the music based on a feature amount of the plurality of music and a coefficient calculated by a multivariate analysis from the impression value of the music. It is characterized by causing a computer to function.

The invention according to claim 74 is based on claim 7.
74. The music search program according to any one of items 0 to 73, wherein the feature amount extracting means causes a computer to function as a feature amount, which is a quantity related to a sounding period and a sounding intensity included in an audio signal of the music piece. I do.

The invention according to claim 75 is the same as that in claim 7.
73. The music search program according to any one of Items 0 to 73, wherein the feature amount extracting means causes a computer to function as a feature amount by calculating a quantity representing a degree of change in the spectrum of an audio signal of the music piece.

The invention according to claim 76 provides the invention according to claim 7
73. The music search program according to any one of Items 0 to 73, wherein the feature amount extracting means causes a computer to function as a feature amount, which is a quantity expressing a sounding frequency of an audio signal of the music piece.

The invention according to claim 77 is the same as that in claim 7
73. The music search program according to any one of items 0 to 73, wherein the feature amount extracting means causes a computer to function as a feature amount by calculating a quantity expressing a degree of sound periodicity of an audio signal of the music piece. I do.

The invention according to claim 78 is based on claim 7
73. The music search program according to any one of Items 0 to 73, wherein the feature amount extracting means causes the computer to function so as to calculate a quantity representing a beat cycle of the music as a feature amount.

The invention according to claim 79 is based on claim 7
73. The music search program according to any one of Items 0 to 73, wherein the feature value extracting means causes the computer to function such that a quantity expressing a beat cycle ratio of the music is calculated as a feature value.

The invention according to claim 80 is based on claim 7
73. The music search program according to any one of Items 0 to 73, wherein the feature value extracting means causes a computer to function so as to calculate, as a feature value, a quantity expressing the beat intensity of the music piece.

The invention according to claim 81 is based on claim 7
73. The music search program according to any one of Items 0 to 73, wherein the feature value extracting means causes the computer to function as a feature value by calculating a quantity representing a ratio of beat intensities of the music.

The invention according to claim 82 is the invention according to claim 7
74. In the music search program according to any one of Items 0 to 73, the feature amount extraction unit divides an audio signal of the music into a plurality of bands, detects a rising component of each band, and uses the rising component of each band. A computer is caused to function so as to calculate a feature amount.

The invention described in claim 83 is the sixth invention.
83. The music search program according to any one of Items 9 to 82, wherein the predicted impression value calculation means calculates the predicted impression value based on a predetermined quantity according to the subjective requirement input by the search requirement input means. The computer is caused to function as described above.

The invention according to claim 84 is based on claim 6.
83. The music search program according to any one of Items 9 to 83, wherein the impression value search means calculates a Euclidean distance between the predicted impression value and an impression value of each music stored in the music database, and The computer is made to function so as to output as candidate songs in order from songs having a short distance.

The invention according to claim 85 is the invention according to claim 6
83. The music search program according to any one of Items 9 to 83, wherein the impression value search means calculates a Euclidean distance between the predicted impression value and an impression value of each music stored in the music database, and For a song whose distance is equal to or less than a predetermined distance, the computer is made to function as a candidate song in order from the song having the smallest Euclidean distance.

The invention according to claim 86 is based on claim 8
86. In the music search program described in 4 or 85, the impression value search means calculates a similarity between the predicted impression value and the impression value of the candidate music based on the Euclidean distance, and outputs the similarity together with the candidate music. The computer is caused to function as described above.

The invention described in claim 87 is the sixth invention.
89. The music selection program according to any one of items 9 to 86, wherein information on the candidate music output by the impression value search means is extracted from the music database and presented to the user, and music is selected from the candidate music. Means for extracting a sound signal of a music piece selected by the music selection means from the database and playing back a part or all of the music piece so that a user can listen to the music piece. It is characterized by the following.

The invention according to claim 88 is the invention according to claim 8
7. The music search program according to 7, wherein the music selection means causes a computer to function so as to select music according to a predetermined music selection device.

The invention described in claim 89 is the same as that in claim 6
90. The music search program according to any one of items 9 to 88, wherein the search requirement input means inputs requirements relating to bibliographic information regarding the music such as a music title, the music database stores bibliographic information regarding the music, and the impression value search. Means for causing a computer to function to search the music database using the bibliographic information as a key.

The invention according to claim 90 is based on claim 8.
9. The music search program according to Item 9, further comprising: bibliographic information input means for inputting bibliographic information on music such as a music title, wherein the music database causes a computer to function so as to store the bibliographic information.

The invention according to claim 91 is based on claim 8
90. The music search program according to 9 or 90, wherein the impression value search means functions the computer so that an impression value of the music corresponding to the bibliographic information input by the search requirement input means is used as a search key as a predicted impression value. It is characterized by making it.

The invention according to claim 92 is the invention according to claim 8.
90. The music search program according to any one of items 9 to 91, wherein the search requirement input means inputs a search device for requirements relating to bibliographic information such as a song title and requirements relating to subjectivity such as an image of a song. A computer is characterized in that a search for bibliographic information and a search for a predicted impression value are performed in combination depending on the device.

The music search data registration program according to claim 93 of the present invention is characterized in that a sound signal input means for inputting a sound signal of a music and a sound signal output from the sound signal input means determine a predetermined characteristic. A computer is made to function so as to include a feature amount extracting unit for extracting an amount, and an impression value calculating unit for calculating an impression value obtained by quantifying the impression of the music from the feature amount.

The invention according to claim 94 is based on claim 9.
3. The music search data registration program according to 3, wherein the feature amount extracting means causes a computer to function as a feature amount to calculate a quantity related to a sound generation cycle and a sound intensity included in an audio signal of the music piece.

The invention according to claim 95 provides the invention according to claim 9
3. The music search data registration program according to item 3, wherein the feature amount extracting means causes a computer to function as a feature amount by calculating a quantity expressing a degree of change in the spectrum of an audio signal of the music piece.

Further, the invention of claim 96 is based on claim 9
3. The music search data registration program according to item 3, wherein the feature amount extracting means causes a computer to function so as to calculate, as a feature amount, a quantity expressing a sounding frequency of an audio signal of the music piece.

The invention according to claim 97 is the same as that in claim 9
3. The music search data registration program according to item 3, wherein the feature value extracting means causes a computer to function as a feature value by calculating a quantity expressing a degree of sound periodicity of an audio signal of the music piece.

Further, the invention according to claim 98 provides the invention according to claim 9
3. The music search data registration program according to item 3, wherein the feature value extracting means causes a computer to function so as to calculate, as a feature value, a quantity expressing a beat cycle of the music piece.

The invention according to claim 99 is the same as that in claim 9
3. The music search data registration program according to item 3, wherein the feature amount extracting means causes a computer to function so as to calculate a quantity expressing a beat cycle ratio of the music as a feature amount.

[0106] The invention according to claim 100 is the music search data registration program according to claim 93, wherein:
The feature amount extracting means causes a computer to function so as to calculate a quantity representing the beat intensity of the music as a feature amount.

The invention according to claim 101 is the music search data registration program according to claim 93, wherein:
The feature amount extracting means causes a computer to function so as to calculate a quantity expressing a ratio of beat intensities of music as a feature amount.

The invention according to claim 102 is a music search data registration program according to claim 93, wherein:
The feature amount extraction unit may divide a sound signal of a musical piece into a plurality of bands, detect a rising component of each band, and cause a computer to calculate a feature amount using the rising component of each band. Features.

Further, in order to achieve the above object, a recording medium according to claim 103 of the present invention, comprises:
Search requirement input means for inputting the user's subjective requirements for the music desired to be searched, digitized and output, and a predicted impression value obtained by quantifying the impression of the search target music from the output of the search requirement input means Predictive impression value calculating means for calculating
A song database that stores sound signals of a plurality of songs and impression values obtained by quantifying the impressions of the songs, and a song search program that functions as an impression value search unit that searches the song database using the predicted impression value as a key is recorded. It is characterized by the following.

The recording medium according to claim 104 of the present invention comprises: an acoustic signal input means for inputting an acoustic signal of a music;
From a sound signal output by the sound signal input means, a feature amount extracting means for extracting a predetermined feature amount, and from the feature amount,
A song search data registration program that causes a computer to function as an impression value calculation unit that calculates an impression value obtained by quantifying the impression of the song is recorded.

The music search method, the music search device and the music search program according to claim 1, 35 and 69 input the subjective requirements of the user for the music desired to be searched, digitize and output. Then, from the output, a predicted impression value is calculated by quantifying the impression of the search target song, and using the predicted impression value as a key, the acoustic signals of the plurality of songs and the impression value obtained by quantifying the impression of the song are stored. By searching the music database described above, it becomes possible to search for a desired music based on a subjective image of the music of the user.

A music search method, a music search device and a music search program according to claims 2, 36 and 70 receive a sound signal of a music and extract a predetermined feature amount from the sound signal. By calculating an impression value that quantifies the impression of the music from the feature amount, it is possible to store the acoustic signal and the impression value in the music database, and to provide a subjective image of the music of the user. , It is possible to search for a desired music.

The music search method, the music search device and the music search program according to claim 3, 37 and 71, input a sound signal of a music to be searched, and determine a predetermined characteristic from the sound signal. By extracting the amount, calculating the impression value of the song from the feature amount, and searching the song database using the calculated impression value as a key, the impression is similar to the subjective impression of the song input as the search target Music can be searched.

The music search method, the music search device and the music search program according to claim 4, claim 38, and claim 72 store the feature amount extracted in the feature extraction step in the music database, and search for the music to be searched. By inputting a search requirement relating to the feature amount of the above, and searching the music database using the requirement relating to the feature amount of the music to be searched as a search key, a music desired by the user is searched based on the feature amount of the music. It becomes possible.

The music search method, the music search device and the music search program according to claim 5, 39 and 73 are calculated by a multivariate analysis or the like from the feature amounts of a plurality of music and the impression values of the music. By calculating the impression value of the music based on the coefficient, it is possible to calculate the impression value of the music necessary for the search based on the subjective image from the feature amount of the music.

The music search method, the music search device and the music search program according to claim 6, 40 and 74 calculate the quantity relating to the sound period and the sound intensity included in the sound signal of the music as the characteristic amount. Accordingly, it is possible to calculate the feature amount necessary for calculating the impression value of the music from the sound signal of the music.

The music search method, the music search device and the music search program according to the seventh to thirteenth, the fourteenth to forty-seventh and the seventy-fifth to eighty-first aspects provide a music search method, a music search apparatus, and a music search program. Calculate the amount of features such as the degree of beat, beat cycle, beat cycle ratio, beat intensity, beat intensity ratio, etc. as the feature amount to calculate the feature amount necessary to calculate the impression value of the song from the audio signal of the song. Becomes possible.

A music search method, a music search device, and a music search program according to claim 14, 48, and 82, divide an audio signal of a music into a plurality of bands, and detect a rising component in each band. By calculating using the rising component of each band, it is possible to calculate the feature amount necessary for calculating the impression value of the music from the sound signal of the music.

The music search method, the music search device and the music search program according to claim 15, claim 49, and claim 83 are arranged so that a predetermined number is set according to the subjective requirements input in the search requirement input step. By calculating the predicted impression value based on the above, it is possible to calculate a predicted impression value necessary for a search for a subjective requirement involving a plurality of subjective factors.

The music search method, the music search device and the music search program according to claim 16, 50 and 84, are arranged so that the estimated impression value and the impression value of each music stored in the music database are stored. By calculating the Euclidean distance of, and outputting the music pieces in order from the one with the smallest Euclidean distance, the candidate music pieces close to the image to be searched can be output in the order of the similarity.

A music search method, a music search device, and a music search program according to claim 17, 51 and 85, wherein the music search method includes a method of setting the predicted impression value and the impression value of each music stored in the music database. By calculating the Euclidean distance of, and outputting the songs whose Euclidean distance is equal to or less than a predetermined distance in order from the music having the smallest Euclidean distance, the candidate music having the similarity within the predetermined range has a large similarity. Can be output in order.

A music search method, a music search device and a music search program according to claim 18, 52 and 86 calculate the similarity between the predicted impression value and the impression value of the candidate music based on the Euclidean distance. Then, by outputting the similarity together with the candidate music, an index of the similarity of the candidate music to the search target can be presented to the user.

The music search method, music search device and music search program according to claim 19, 53 and 87 extract information on the candidate music output by the impression value search step from the music database. Presented to the user, allowing the user to select a song from the candidate songs, extracting the audio signal of the song selected by the user from the database, and allowing the user to listen to part or all of the song By replaying the search result, the user can preview and listen to the search result, and can narrow down the search result.

The music search method, the music search device and the music search program according to claim 20, 54 and 88 select a plurality of candidate music by selecting music to be extracted from the candidate music in accordance with a predetermined music selection method. You can narrow down the songs to be output from the songs.

The music search method, the music search device and the music search program according to the twenty-first, twenty-fifth and fifty-ninth aspects input requirements relating to bibliographic information on music such as a music title, and use the bibliographic information as a key for bibliographic information. By searching the music database in which information is stored, it is possible to perform a search based on bibliographic information of music such as a music title, in addition to a search based on a subjective image of music.

A music search method, a music search device and a music search program according to claims 22, 56 and 90 input bibliographic information on music such as music titles and store the bibliographic information in the music database. By doing
Bibliographic information required for bibliographic search can be registered in the music database.

The music search method, music search device and music search program according to claim 23, claim 57 and claim 91, input information relating to the bibliographic information of the music to be searched, and enter the bibliographic information into the input bibliographic information. By using the impression value of the music to be performed as a search key as the predicted impression value, it is possible to search for music whose image is subjectively similar to the music related to the bibliographic information input by the user.

The music search method, the music search device and the music search program according to the twenty-fourth, fifty-eighth, and ninety-second input a search method for a requirement relating to bibliographic information such as a song title and a requirement relating to subjectivity such as a music image. Then, by performing a search on the bibliographic information in combination with a search on the predicted impression value in accordance with the search method, it is possible to search for the desired music by combining the bibliographic information and the subjective image of the search target music. Becomes

A music search data registration method, a music search data registration device, and a music search data registration program according to claims 25, 59, and 93, input a sound signal of a music and read the sound signal from the sound signal. By extracting a predetermined feature amount and calculating an impression value by quantifying the impression of the music from the feature amount,
The sound signal and the impression value can be registered.

A music search data registration method, a music search data registration device, and a music search data registration program according to claim 26, claim 60, and claim 94, comprise a sound generation cycle and a sound intensity included in an audio signal of a music. By calculating the quantity related to the music as a feature quantity, it is possible to calculate the feature quantity necessary for calculating the impression value of the music from the sound signal of the music.

Claims 27 to 33 and Claims 61 to 67
A music search data registration method, a music search data registration device, and a music search data registration program according to claims 95 to 101, wherein the degree of change in the spectrum of the sound signal of the music, the pronunciation frequency, the pronunciation periodicity, the beat cycle By calculating the quantities such as the beat cycle ratio, the beat intensity, and the beat intensity ratio as the feature amount, it is possible to calculate the feature amount necessary for calculating the impression value of the music from the audio signal of the music.

A music search data registration method, a music search data registration device, and a music search data registration program according to claims 34, 68, and 102, divide an audio signal of a music into a plurality of bands and start each band. Detect the component,
By calculating using the rising component of each band, it is possible to calculate the feature amount necessary for calculating the impression value of the music from the sound signal of the music.

[0133]

(Embodiment 1) Hereinafter, Embodiment 1 of the present invention will be described with reference to the drawings.
FIG. 1 is a block diagram showing an overall configuration of a music search device according to Embodiment 1 of the present invention. In FIG. 1, 11 is a music database, 12 is an audio signal input unit, 13 is a feature amount extraction unit, 14 is an impression value calculation unit, 15 is a bibliographic information input unit, 16 is a search requirement input unit, and 17 is a predicted impression value calculation. Means, 18 means impression value searching means, 19 means music selection means, 1
Reference numeral 10 denotes a music reproducing means.

The operation of the music search apparatus configured as described above will be described below with reference to FIG. This device is roughly divided into a music registration unit 111 for registering an audio signal of a music to be searched and its accompanying information in the music database 11, and a music for searching a desired music from the music database according to a user's search requirements. It comprises a search unit 112 and a music database 11.

First, the music registration section 111 will be outlined.
The music database 11 records, for each music piece, a plurality of music pieces that can be reproduced by the user, and records the sound signals and accompanying information for each music piece so that the music piece data can be referred to. First, the sound signal input unit 12 outputs the sound signal (CD
The audio signal recorded on a medium such as an audio signal and the audio signal distributed in a broadcast or the like) are registered in the music database 11 and output to the feature extraction means 13 in the subsequent stage for generating accompanying information. If the input audio signal is an analog signal,
After being digitized by the audio signal input means 12, it is output to the subsequent stage.

Next, the feature value extracting means 13 extracts some feature values representing the physical features of the input audio signal from the input audio signal, and registers them in the music database. Then, the impression value calculation unit 14 converts the physical feature amount of the music extracted by the feature amount extraction unit 13 into an impression value obtained by quantifying the subjective impression of the music, and registers the impression value in the music database. On the other hand, bibliographic information (song name, singer name, etc.) related to the audio signal input as a registration target is separately registered in the music database by the bibliographic information input unit 15.

Next, the music search section 112 will be outlined.
First, the user inputs the requirements for the music to be searched into the search requirement input means 16. Next, the predicted impression value calculation means 17 calculates an impression value (a quantity representing a subjective impression of the music) of the music expected as a search target from a part related to the subjective impression in the input search requirement. Then, the impression value search means 18 refers to the accompanying information part of each music registered in the music database 11 using the input search requirement and the expected impression value calculated by the predicted impression value calculation means 17 as a search key, and performs a search. A plurality of music pieces having accompanying information similar to the key are output as candidate music pieces. Thereafter, the music selecting means 19 selects a music by a user's selecting operation or a predetermined music selecting procedure, and the music reproducing means 110 reproduces the finally selected music as a search result.

The details of each means will be described below with reference to the drawings. FIG. 2 is a block diagram schematically showing a feature amount extracting / extracting unit according to the present embodiment. Table 1 shows feature amounts and an outline thereof in the embodiment of the present invention.

[0139]

[Table 1]

The feature extracting means of FIG. 2 extracts eight physical features described in (Table 1) from the input acoustic signal. The operation of extracting each feature amount in (Table 1) will be described with reference to FIG. Hereinafter, it is assumed that the processing frame time is t, the frequency band number obtained by band division is f, and the audio signal is digitized and then processed in units of a frame having a fixed time length.

Spectral change (SF) First, in step 21, the input acoustic signal is Fourier-transformed for each processing frame, and the power of each band is obtained to calculate the power spectrum S (t). Next, in step 22, the power spectrum variation ΔS (t) between frames is calculated from (Equation 1).

[0142]

(Equation 1)

Then, in step 23, ΔS (t) is averaged over all frame times, and the spectral change S
FLX is calculated (Equation 2).

[0144]

(Equation 2)

Average Number of Pronunciations (AR) First, the power spectrum S calculated in step 21
Using the power p (t, f) of each band of (t), the rising degree d (t, f) and d of the signal component of each band is used.
The rising component D (t) is calculated by adding (t, f) in the frequency direction at the frame time t. d (t, f)
And D (t) are calculated from (Equation 3) to (Equation 7) using the power p (t, f) of each band f.

[0146]

(Equation 3)

[0147]

(Equation 4)

[0148]

(Equation 5)

[0149]

(Equation 6)

[0150]

(Equation 7)

A method of extracting the rising degree d (t, f) and the rising component D (t) is described in, for example, a literature (Goto and Muraoka: “Beat tracking system for music acoustic signal”, Information Processing Society of Japan Research Report, Vol. 94). , No.7
1, pp. 49-56, 1994). Then, in step 25, the appearance frequency of the rising degree d (t, f) over all frames is calculated by (Equation 8),
The average number of pronunciations.

[0152]

(Equation 8)

First, in step 26, the autocorrelation function A (m) (m is a delayed frame) of D (t) is calculated by (Equation 9) in step 26, and the periodicity of the rising component is obtained. Next, in step 27, A (m) is Fourier-transformed and the power of each band is obtained to obtain a power spectrum Aspect of A (m).
(K) (K is a frequency) is calculated. And step 28
At the DC component value of Aspect (K) (ie Aspe (K)
c (0)) in step 29 with Aspect (K)
(Specified as Spec (Kpeak)), and in step 210 Aspect (0)
By calculating the ratio of Aspec (Kpeak), the sound non-periodicity is calculated (Equation 10).

[0154]

(Equation 9)

[0155]

(Equation 10)

Beat period (TT) The beat period TT is the interval between the beats of the music, that is, the time length of the quarter note of the music, and is calculated using the autocorrelation function A (m) of the rising component D (t). To detect. First, step 211
Then, the time length pk (i) in which the peak portion of A (m), that is, the period of the rising component of the signal appears most strongly is calculated. Next, in step 212, beat cycle candidates T1 and T2 are calculated from pk (i), the periodic structure of the music is calculated in step 213, and the rising frequency A is calculated in step 214.
R, and by referring to the periodic structure, T1,
The one considered to be the true beat cycle of T2 is output as the beat cycle of the music. For the calculation method of the beat cycle, see the literature (Tagawa and Misaki: “A study on tempo detection from music signals”, Proc. Of the Acoustical Society of Japan, pp.529-530, 200
Details are described in (0).

Beat cycle ratio (BR) The beat cycle ratio is determined from the relationship between the beat cycle and the dominant sounding cycle. First, in step 215, the Aspect
After calculating the time period Tkpeak corresponding to (Kpeak), the ratio BR between the beat period TT and Tkpeak is calculated by (Equation 11).

[0158]

[Equation 11]

Beat Intensity 1 (BI1) Calculates the power of the rising component that appears in approximately half the cycle of the beat. First, in step 216, a frequency f1 corresponding to a half of the beat cycle is calculated from TT. Then, in step 217, As having the maximum value near f1
The value of ped (K) is referred to, and the value is set to BI1.

Beat Intensity 2 (BI2) Calculates the power of the rising component that appears in a cycle approximately one-fourth of the beat cycle. First, in step 218, a frequency f2 corresponding to half of the beat cycle is calculated from TT. Then, in step 219, A having the maximum value near f2
The value of spec (K) is referred to, and the value is set to BI2.

Beat intensity ratio (IR) In step 220, the beat intensity 1BI1 and the beat intensity 2B
The ratio IR of I2 is calculated by (Equation 12).

[0162]

(Equation 12)

Each of the above feature amounts is a quantity representing an acoustic feature of the music, and is considered to be highly relevant to the subjective impression given to the listener by the music. For example, the beat period is a quantity that expresses the tempo of a song, that is, the speed. Generally, a fast song feels "busy" and a slow song feels "slow". That is. Therefore, these features are used as quantities for associating with a subjective impression.

The feature values (Table 1) obtained by the feature value extracting means 13 in FIG.
The impression value, that is, the quantity expressing the subjective impression of the music, is converted by (Equation 13).

[0165]

(Equation 13)

In Equation (13), Ii is the impression value for the impression factor i, Pj is the value of the feature amount j, and Wij is the feature amount j
Np is the number of types of the feature amount, and Np is a weighting coefficient indicating the relationship between the feature amount and the impression factor i. In the present embodiment, Np = 7 as shown in (Table 1), and Pj corresponds to each feature amount. The impression value Ii is a quantity representing the subjective impression of the music, and is used as a quantity obtained by summing up the degree of impression (Ej) represented by a specific adjective. For example, the impression of the music is “intensity (E1)”, “goodness of the paste (E2)”, “freshness (E3)”, “simpleness (E
If expressed in terms of the five viewpoints of "4)" and "softness (E5)", Ii is obtained from Ej by (Equation 14).

[0167]

[Equation 14]

In equation (14), Yij is a weighting factor representing the relationship between Ej and Ij. Derivation of Yij and the number of impression factors Ni is based on Ej for a plurality of music pieces selected as samples.
Is determined in advance by a sensory evaluation experiment using the SD (semantic differencial) method or the like, and the result is subjected to factor analysis such as principal component analysis. Also, W
The derivation of ij is based on the sensory evaluation and factor analysis described above.
, And the impression value Ii of each song is calculated by (Equation 14).
The relationship with the feature amount Pj calculated for each corresponding song is calculated by a method such as linear multiple regression analysis. Note that the feature amount Pj and the impression value Ii may be derived by a nonlinear system such as a neural network instead of the linear multiple regression analysis.

The music database 11 is an aggregate of records composed of audio signals of each music and its accompanying information. FIG. 3 shows an example of a record configuration of the music database 11 in the present embodiment. The record includes a song ID for uniquely referring to the record, bibliographic information on the song such as a song title and a singer name input by the bibliographic information input unit 15, a feature amount extracted by the feature amount extraction unit 13, a feature amount → an impression It is composed of the impression value extracted by the value conversion means and the sound signal of the music input by the sound signal input means 12.

[0170] The search requirement input means 16 inputs the requirements for the music that the user desires to search. FIG. 4 shows an example of the input search requirement. As the search requirements, a character string such as an artist name, a numerical value indicating the degree of "intensity", and the presence or absence of a predetermined condition such as "I want to dance cheerfully" and a search method are input. Each search requirement is input by a user using a keyboard, various switches, a slider, a volume, or the like.

[0171] The predicted impression value calculation means 17 calculates the impression value PIj () of the music expected as a search target from the requirements relating to the subjective impression (subjective requirements) among the search requirements input by the search requirement input means 16. (Predicted impression value). The predicted impression value PIj is calculated from Expression 15 using the weight coefficient Yij.

[0172]

(Equation 15)

IEj is a numerical value that constitutes a subjective requirement such as the degree of "intensity", and the type of IEj corresponds to the type of the degree of impression on the music used in the process of deriving the weight coefficient Yij.

For requirements involving a plurality of subjective factors, such as “I want to dance cheerfully”, the value of IEj expressing the requirement is preset, and if the requirement is selected, the preset value is set. From (Equation 15), a predicted impression value PIj is calculated. For example, when the subjective degree is expressed by a quantity between the maximum of 1.0 and the minimum of -1.0, the requirement of "want to dance cheerfully" is analogized to "the degree of goodness and exhilaration is great." Because it is possible, intensity IE1 =
0.5, good IE2 = 1.0, freshness IE3 =
1.0, simple IE4 = 0.0, soft IE5 = 0.
Preset by a value such as 0, a corresponding predicted impression value PIj can be calculated from IEj.

The impression value searching means 18 extracts a record similar to the search key using the bibliographic requirement and the predicted impression value PIj as a search key with reference to the music database 11. For bibliographic requirements, by referring to the bibliographic information part of each record stored in the music database, by matching with the input requirements (whether or not the input character strings are included), Extract similar records.
On the other hand, for the predicted impression value PIj, the similarity with the impression value of each record stored in the music database is quantified and searched. FIG. 5 schematically illustrates the similarity of impression values. The impression value Ii of each music piece recorded in each record can be expressed in space as a vector composed of a number set of Ni number of impression factors, and this space is called an impression space. FIG. 5 shows the number N of impression factors of the impression value.
This is an example where i = 2, and the impression value Ii of each music piece is represented by a point on a two-dimensional impression space, such as 42. Similarly, it is assumed that the predicted impression value PIj can also be laid out in the impression space, and laid out at a position 43 as an example for a certain subjective requirement. Then, the predicted impression value PIj and the music impression value Ii
Can be expressed as the magnitude of the Euclidean distance on the impression space, and the distance L can be calculated and quantified by (Equation 16).

[0176]

(Equation 16)

When the distance L is calculated for all songs to be searched, songs with smaller L can be regarded as more similar to the predicted impression value, that is, similar to the subjective impression of the desired song. The music pieces having short distances L are sequentially set as candidate music pieces. As the candidate music output as the search result, a predetermined number of candidate music is output as the search result. Note that 44
, A similarity determination boundary centering on the predicted impression value may be set, and the music within the boundary may be output as a search result. In this case, by defining a similarity determination range in advance, it is possible to add a limitation that music pieces having a similarity degree of a certain degree or more are not candidate music pieces.

Further, the search based on the bibliographic requirements and the search based on the subjective requirements may be performed independently or in combination. The search method is specified by the user using the search requirement input means.

[0179] Further, the bibliographic information such as the title of the song input by the search requirement input means 16 is searched for the above-mentioned song first, and the song is searched for using the impression value of the song as the predicted impression value. Based on this, it is possible to search for music that is subjectively similar. For example, the user inputs "Night Sky Nomuko" as a song title, and first searches the song database 11 using the song title as a key. After the hit song is found, a search is performed based on the subjective impression of the song by further searching the song database 11 using the impression value of “Night Sky Nomuko” as the predicted impression value, and a similar impression is obtained. Can be searched for. In this example, “song having an impression similar to“ Night Sky Nomuko ”” is output as a search result.

When the search requirement input means 16 includes the audio signal input means 12, the feature quantity extraction means 13, and the impression value calculation means 14, the feature quantity is calculated from the input sound signal, and the feature quantity is calculated. By referring to the music database as the predicted impression value, it is possible to search for music having an impression similar to the sound signal of the input music.

The candidate songs extracted by the impression value search unit 18 can be further narrowed down by the song selection unit 19. The music selecting means 19 extracts accompanying information (song name, artist name, etc.) relating to the extracted candidate music from the music database and presents it to the user, thereby prompting the user to select a music to be reproduced. At the time of selection, a part or all of the music may be viewed by the music reproducing means 110.

When the search is performed according to the subjective requirement, based on the distance L calculated by the impression value search means 18,
The similarity between the subjective requirement specified by the user and each of the extracted candidate songs may be calculated, and the similarity may be presented to the user. Note that the selection of the search result may be automatically performed without a user's instruction by a predetermined method such as “the first candidate song is selected”. Presentation of information to the user is performed by display means such as a monitor or LCD, and selection input is performed by input means such as a keyboard or a switch.

The music selected by the music selecting means is presented by the music reproducing means 110 in a form that the user can listen to. Of course, without reproducing the music, bibliographic information such as the music title may be presented to the user as a search result.

(Embodiment 2) Hereinafter, Embodiment 2 of the present invention will be described with reference to the drawings. FIG. 6 is a block diagram showing an overall configuration of a music search data registration device according to Embodiment 2 of the present invention. In FIG. 6, 6
1 is a music database, 62 is an audio signal input means, 63
Denotes a feature amount extracting unit, 64 denotes an impression value calculating unit, and 65 denotes a bibliographic information input unit. This device is a device in which the music registration unit 111 according to the first embodiment is independent. Therefore, the operation of each means in FIG. 6 is the same as each means described in the first embodiment.

The music search data registration device of the present embodiment extracts music to be registered in the music database as search data and its accompanying information from the input sound signal, and stores the extracted music in the music database 61. The data stored in the database 61 by the present device is an acoustic signal, a feature value, an impression value, bibliographic information, and the like. By storing this data in the music database, it is necessary for the music search device according to the first embodiment. Databases can be created individually.

The music database 61 is in the form of a recording medium such as a hard disk or a removable medium such as a CDROM. After the music data is registered by the present apparatus, it can be incorporated into a music search device and used for searching.

In some cases, the music database 61 and other means are connected via a network such as the Internet. In this case, after the registration data is created by the present apparatus, the registration data is stored in the database via the network. By storing, one connected to the network
Alternatively, data necessary for a search can be supplied to a plurality of music search devices.

(Embodiment 3) This embodiment is an embodiment in which Embodiment 1 and Embodiment 2 are implemented as a computer program. Hereinafter, a third embodiment of the present invention will be described with reference to the drawings. FIG. 7 is a block diagram showing an overall configuration of a music search program according to Embodiment 3 of the present invention. In FIG. 7, reference numeral 71 denotes a music registration program, 72 denotes a music search program, 73 denotes a music database, and the names and functions of other means are the same as those in the first embodiment.

The music registration program 71 and the music search program 72 are described as programs for various computers such as a personal computer and a microcomputer, and are stored in a recording area (memory, hard disk, floppy (registered trademark) disk, etc.) of the computer. ing. Further, as described in the first and second embodiments, the music database 73 stores a hard disk or a CDRO.
It is an aggregate of data recorded on a computer recording medium such as M.

The music registration program 71 extracts music to be registered in the music database as search data and its accompanying information from the audio signal input to the computer, and stores it in the music database 73. The data stored in the music database 73 by this program is an acoustic signal, a feature value, an impression value, bibliographic information, and the like. By storing this data in the music database 73, a database necessary for the music search program 72 is created. can do.

On the other hand, the music search program 72 first inputs a user's search requirement, and obtains an impression value (predicted impression value) of a music expected as a search target from a part relating to a subjective impression in the search requirement. calculate. Then, by using the input search requirement and the predicted impression value as search keys, referring to the accompanying information portion of each song registered in the song database 73, a plurality of songs having accompanying information similar to the search key as candidate songs. Output. After that, music is selected according to a selection parameter or a predetermined music selection procedure that symbolizes the user's selection operation, and the finally selected music is output as a search result.

According to the above-mentioned program, it is possible to provide a desired music in accordance with the search requirement input by the user. The music registration program 71 and the music search program 72 may be stored in a recording area of a computer via a removable medium such as a CDROM 91 (FIG. 9), a computer network, or the like. The music registration program 71 and the music search program 72 may operate on different computers, respectively, and access a music database 73 stored in a recording area shared by both computers. Of course, the music database 73 is recorded on a removable medium such as a floppy disk or CDROM, and the music
3 may be used by the music search program 72 running on another computer.

[0193]

According to the inventions described in claims 1, 35 and 69, the subjective requirements of the user for the music desired to be searched are input and quantified and output.
A predicted impression value obtained by quantifying the impression of the search target song is calculated, and using the predicted impression value as a key, the music database that stores the sound signals of a plurality of songs and the impression value obtained by quantifying the impression of the song is searched. By doing so, it is possible to search for a desired music piece based on the user's subjective image of the music piece.

According to the invention of claims 2, 36 and 70, an audio signal of a music is inputted, a predetermined characteristic amount is extracted from the audio signal, and an impression of the music is extracted from the characteristic amount. By calculating the impression value quantifying
The sound signal and the impression value can be stored in the music database, and a desired music can be searched based on a subjective image of the music of the user.

According to the invention of claims 3, 37 and 71, an audio signal of a music to be searched is input, a predetermined characteristic amount is extracted from the audio signal, and a predetermined characteristic amount is extracted from the characteristic amount. By calculating the impression value of the song and searching the song database using the calculated impression value as a key,
It is possible to search for music that is similar to the subjective impression of the music input as the search target.

According to a fourth aspect of the present invention, the feature amount extracted in the feature extracting step is stored in the song database, and a search requirement relating to the feature amount of the search target song is input. By searching the music database using the requirement regarding the feature amount of the search target music as a search key, it is possible to search for a music desired by the user based on the feature amount of the music.

The invention according to claim 5, 39, or 73 is characterized in that an impression value of a music piece is determined based on a feature amount of a plurality of music pieces and a coefficient calculated by a multivariate analysis from the impression value of the music piece. The calculation makes it possible to calculate the impression value of the music necessary for the search based on the subjective image from the feature amount of the music.

According to the sixth, forty and forty-fourth aspects of the present invention, the quantity relating to the sound generation period and the sound intensity included in the sound signal of the musical composition is calculated as a feature quantity.
It is possible to calculate the feature amount necessary for calculating the impression value of the music from the sound signal of the music.

The invention according to claims 7 to 13, 41 to 47 and 75 to 81 is characterized in that the degree of change in the spectrum of the sound signal of the music, the frequency of sound, the degree of sound periodicity, the beat cycle, the beat cycle ratio, By calculating the quantity such as the pulse strength and the pulse strength ratio as the feature quantity, it is possible to calculate the feature quantity necessary for calculating the impression value of the music from the sound signal of the music.

According to the invention of claims 14, 48 and 82, the sound signal of the music is divided into a plurality of bands, the rising components of each band are detected, and the calculation is performed using the rising components of each band. By doing so, it becomes possible to calculate the feature amount necessary for calculating the impression value of the music from the sound signal of the music.

According to the fifteenth, fifty-ninth, and eighteenth aspects of the present invention, a predicted impression value is calculated based on a predetermined quantity according to a subjective requirement input in the search requirement input step. For a subjective requirement involving a plurality of subjective factors, a predicted impression value required for a search can be calculated.

The invention according to claim 16, 50, or 84 calculates the Euclidean distance between the predicted impression value and the impression value of each music stored in the music database, and calculates the Euclidean distance. By outputting music pieces in order from the smallest music piece, candidate music pieces that are close to the image to be searched can be output in descending order of similarity.

The invention according to claim 17, 51 or 85 calculates the Euclidean distance between the predicted impression value and the impression value of each music stored in the music database, and calculates the Euclidean distance. By outputting the music pieces whose distance is equal to or less than the predetermined distance in order from the music piece having the smallest Euclidean distance, the candidate music pieces having the similarity within the predetermined range can be output in the order of the similarity.

According to the inventions of claims 18, 52 and 86, the similarity between the predicted impression value and the impression value of the candidate music is calculated based on the Euclidean distance, and the similarity is output together with the candidate music. By doing so, it is possible to present to the user an index of the degree of similarity of the candidate song to the search target.

According to the invention of claims 19, 53 and 87, the information on the candidate music output by the impression value search step is extracted from the music database and presented to the user. Allows the user to select a song from the candidate songs, extracts the audio signal of the song selected by the user from the database, and reproduces part or all of the song so that the user can listen to it, so that the user can search. The results can be checked by trial listening, and the search results can be narrowed down.

According to the twentieth, fifty-fourth, and eighteenth aspects of the present invention, it is possible to narrow down the music to be output from a plurality of candidate music by selecting music to be extracted from the candidate music in accordance with a predetermined music selection method. .

According to the twenty-first, fifty-fifth, and eighty-ninth aspects of the present invention, a requirement relating to bibliographic information relating to a song such as a song title is input, and the song database in which bibliographic information is stored is searched using the bibliographic information as a key. This makes it possible to perform a search based on bibliographic information of a song such as a song title, in addition to a search based on a subjective image of the song.

According to the invention of claims 22, 56 and 90, bibliographic information relating to music such as music titles is inputted,
By storing the bibliographic information in the music database, bibliographic information required for bibliographic search can be registered in the music database.

According to the twenty-third, fifty-seventh, and nineteenth aspects of the present invention, information relating to the bibliographic information of a music to be searched is input, and the impression value of the song corresponding to the input bibliographic information is set as a predicted impression value. By using the search key, it is possible to search for music whose image is subjectively similar to the music related to the bibliographic information input by the user.

According to the invention of claims 24, 58 and 92, a method of searching for a requirement relating to bibliographic information such as a song title and a requirement relating to subjectivity such as an image of a musical composition is inputted, and bibliographic information is entered according to the retrieval method. By performing a search in combination with a search on the predicted impression value, the bibliographic information and the subjective image of the music to be searched are combined,
It is possible to search for a desired music.

According to the twenty-fifth, fifty-fifth, and ninety-third aspects of the present invention, an audio signal of a music is input, a predetermined characteristic amount is extracted from the audio signal, and an impression of the music is extracted from the characteristic amount. By calculating the impression value quantifying
The sound signal and the impression value can be registered in a music database.

According to the twenty-sixth, sixty, and ninety-fourth aspects of the present invention, an impression value of a musical piece is calculated by calculating as a feature quantity a quantity related to a sound generation period and a sound intensity included in an audio signal of the music piece. Can be calculated from the audio signal of the music.

Claims 27 to 33 and Claims 61 to 67
The invention according to claims 95 to 101 is characterized in that the quantities such as the degree of change in the spectrum of the sound signal of the music, the frequency of sounding, the degree of sounding periodicity, the beat cycle, the beat cycle ratio, the beat intensity, the beat intensity ratio, etc. By calculating as, it is possible to calculate the feature amount necessary for calculating the impression value of the music from the sound signal of the music.

According to the invention of claims 34, 68 and 102, the sound signal of the music is divided into a plurality of bands, the rising component of each band is detected, and the calculation is performed using the rising component of each band. Accordingly, it is possible to calculate the feature amount necessary for calculating the impression value of the music from the sound signal of the music.

According to the invention of claim 103, the user's subjective requirements for the music desired to be searched are input, quantified and output, and from the output, the impression impression of the search target music is quantified. A program for calculating a value and using the predicted impression value as a key is recorded in a recording medium, the program functioning to search the music database storing sound signals of a plurality of music and impression values obtained by quantifying impressions of the music. Thereby, the desired music can be searched for based on the subjective image of the music of the user by the computer program recorded on the recording medium.

[0216] According to the invention, an audio signal of a music is inputted, a predetermined characteristic amount is extracted from the audio signal, and an impression value quantifying the impression of the music is calculated from the characteristic amount. By recording a program that causes the computer to record the sound signal and the impression value in a music database using a computer program recorded on the recording medium.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a schematic configuration of a music search device according to a first embodiment of the present invention.

FIG. 2 is a block diagram showing a schematic configuration of feature amount extraction according to the first embodiment of the present invention;

FIG. 3 is a diagram showing an example of records constituting a music database according to the first embodiment of the present invention;

FIG. 4 is a diagram illustrating an example of an input requirement according to the first embodiment of the present invention;

FIG. 5 is a diagram showing a schematic description of an impression space.

FIG. 6 is a block diagram illustrating a schematic configuration of a music search data registration device according to a second embodiment of the present invention.

FIG. 7 is a block diagram illustrating a schematic configuration of a music search program according to a third embodiment of the present invention.

FIG. 8 is a block diagram showing a schematic configuration of a conventional music search device.

FIG. 9 is a diagram showing a CDROM according to a third embodiment of the present invention.

[Explanation of symbols]

 11, 61, 73, 82 Music database 12, 62 Acoustic signal input means 13, 63 Feature amount extraction means 14, 64 Impression value calculation means 15, 65 Bibliographic information input means 16 Search requirement input means 17 Predicted impression value calculation means 18 Impression Value search means 19 Music selection means 31 Music record 32 Bibliographic information 33 Feature value 34 Impression value 71 Music registration program 72 Music search program 81 Music selection condition input means 83 Bibliographic information search means 84 Playback music selection means 85,110 Music reproduction means 91 CDROM 111 Music registration section 112 Music search section

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat ゛ (Reference) G10L 11/00 G10L 9/00 A // G10L 15/10 3/00 531N (72) Inventor Sansaki ▼ Masayuki 1006 Kazuma Kadoma, Kadoma City, Osaka Prefecture F-term (reference) in Matsushita Electric Industrial Co., Ltd. 5B075 ND14 ND34 NK06 NK12 NK43 PP02 PP12 PP24 PQ04 PQ36 PR06 QM08 UU37 5D015 HH23 5D082 BB01 5D378 XX43

Claims (104)

[Claims] 1. A search requirement inputting step of inputting a user's subjective requirement for a music desired to be searched, digitizing and outputting the result, and obtaining an impression of the music to be searched from the output of the search requirement inputting step. A predicted impression value calculating step of calculating a converted predicted impression value; a music database storing sound signals of a plurality of songs and an impression value quantifying the impression of the song; and a music database using the predicted impression value as a key. A music search method comprising: an impression value search step for searching. 2. An audio signal inputting step of inputting an audio signal of a music piece; a feature quantity extracting step of extracting a predetermined feature quantity from the audio signal; and quantifying an impression of the music piece from the feature quantity. The music search method according to claim 1, further comprising an impression value calculation step of calculating an impression value, wherein the music database stores the acoustic signal and the impression value. 3. The search requirement input step includes a music impression value calculation step of calculating an impression value of the input music, and the music impression value calculation step inputs an audio signal of a music to be searched. An audio signal inputting step; a feature quantity extracting step of extracting a predetermined feature quantity from the acoustic signal; and an impression value calculating step of calculating an impression value of a music piece from the feature quantity. 3. The music search method according to claim 1, wherein the music database is searched using the impression value calculated in the value calculation step as a key. 4. The music database stores the feature amount extracted in the feature extraction step, the search requirement input step inputs a search requirement related to a feature amount of a search target music, and the impression value search step 3. The music search method according to claim 2, wherein the music database is searched using a requirement relating to a feature amount of the search target music as a search key. 5. The method according to claim 1, wherein the impression value calculating step calculates an impression value of the music based on a feature amount of a plurality of music and a coefficient calculated by a multivariate analysis from the impression value of the music. The music search method according to claim 2. 6. The feature amount extracting step according to claim 2, wherein the feature amount extracting step calculates, as a feature amount, a quantity relating to a sound generation period and a sound intensity included in the sound signal of the music piece. Music search method. 7. The music search according to claim 2, wherein the feature amount extraction step calculates, as a feature amount, a quantity expressing a degree of change in the spectrum of an audio signal of the music. Method. 8. The music search method according to claim 2, wherein the feature amount extraction step calculates, as a feature amount, a quantity expressing a sounding frequency of an audio signal of the music. . 9. The feature amount extraction step according to claim 2, wherein the feature amount extraction step calculates, as the feature amount, a quantity expressing the degree of the sound periodicity of the audio signal of the music piece. Music search method. 10. The music search method according to claim 2, wherein the feature amount extraction step calculates a quantity representing a beat cycle of the music as a feature amount. 11. The music search method according to claim 2, wherein the feature amount extraction step calculates a quantity expressing a beat cycle ratio of the music as a feature amount. 12. The music search method according to claim 2, wherein the feature amount extraction step calculates a quantity expressing the beat intensity of the music as a feature amount. 13. The music search method according to claim 2, wherein the feature amount extraction step calculates a quantity expressing a ratio of beat intensities of the music as a feature amount. 14. The feature amount extracting step includes dividing a sound signal of the music into a plurality of bands, detecting a rising component of each band, and calculating a feature amount using the rising component of each band. The music search method according to any one of claims 2 to 5, wherein 15. The method according to claim 1, wherein the predictive impression value calculating step calculates a predictive impression value based on a predetermined quantity according to the subjective requirement input in the search requirement input step. 15. The music search method according to any one of items 1 to 14. 16. The impression value search step calculates a Euclidean distance between the predicted impression value and an impression value of each song stored in the song database, and selects candidate songs in order from the song with the smallest Euclidean distance. The music search method according to any one of claims 1 to 15, wherein the music search result is output as the following. 17. The impression value searching step calculates a Euclidean distance between the predicted impression value and an impression value of each music stored in the music database, and the Euclidean distance is equal to or less than a predetermined distance. The music search method according to any one of claims 1 to 15, wherein the music is output as candidate music in order from the music having the smallest Euclidean distance. 18. The impression value searching step calculates a similarity between the predicted impression value and an impression value of a candidate song based on the Euclidean distance, and outputs the similarity together with the candidate song. The music search method according to claim 16. 19. A music selection step of extracting information about the candidate music output from the impression value search step from the music database and presenting the information to a user, and selecting music from the candidate music; 19. A music reproduction step of extracting an audio signal of the performed music from the database and reproducing part or all of the music so that a user can listen to the music. Music search method described in section. 20. The music selection step according to claim 1, wherein the music selection is performed according to a predetermined music selection method.
9. The music search method according to item 9. 21. The search requirement inputting step inputs requirements relating to bibliographic information on music such as a song title, the music database stores bibliographic information on music, and the impression value searching step uses the bibliographic information as a key. The music search method according to any one of claims 1 to 20, wherein the music database is searched. 22. The music search method according to claim 21, further comprising a bibliographic information input step of inputting bibliographic information on music such as a music title, wherein the music database stores the bibliographic information. 23. The impression value search step, wherein the impression value of the song corresponding to the bibliographic information input in the search requirement input step is used as a search key as a predicted impression value. Music search method described. 24. The search requirement input step inputs a search method for a requirement relating to bibliographic information such as a song title and a requirement relating to subjectivity such as an image of a song, and the impression value retrieval step performs a search relating to bibliographic information according to the search method. The music search method according to any one of claims 21 to 23, wherein the search is performed in combination with a search on a predicted impression value. 25. An audio signal inputting step of inputting an audio signal of a musical composition, a characteristic amount extracting step of extracting a predetermined characteristic amount from an audio signal output by the audio signal inputting step, A music search data registration method, comprising: an impression value calculation step of calculating an impression value obtained by quantifying a music impression. 26. The music search data registration method according to claim 25, wherein the feature amount extracting step calculates, as a feature amount, a quantity relating to a sounding period and a sounding intensity included in an audio signal of the music piece. 27. The music search data registration method according to claim 25, wherein the feature quantity extraction step calculates a quantity expressing a degree of change in the spectrum of an audio signal of the music piece as a feature quantity. 28. The music search data registration method according to claim 25, wherein, in the feature amount extracting step, a quantity expressing a sounding frequency of an audio signal of the music piece is calculated as a feature amount. 29. The music search data registration method according to claim 25, wherein, in the feature amount extraction step, a quantity expressing a degree of a sound periodicity of an audio signal of the music piece is calculated as a feature amount. 30. The music search data registration method according to claim 25, wherein the feature amount extracting step calculates a quantity expressing a beat cycle of the music as a feature amount. 31. The music search data registration method according to claim 25, wherein the feature quantity extraction step calculates a quantity representing a beat cycle ratio of the music piece as a feature quantity. 32. The music search data registration method according to claim 25, wherein the feature amount extracting step calculates a quantity representing the beat intensity of the music as a feature amount. 33. The music search data registration method according to claim 25, wherein the feature quantity extraction step calculates a quantity representing a ratio of beat intensities of the music pieces as a feature quantity. 34. The feature amount extracting step includes dividing a sound signal of the music into a plurality of bands, detecting a rising component in each band, and calculating a feature amount using the rising component in each band. 26. The music search data registration method according to claim 25, wherein 35. A search requirement input means for inputting a user's subjective requirement for a music desired to be searched for and digitizing and outputting the numerical value. A predicted impression value calculating means for calculating a converted predicted impression value; a music database storing sound signals of a plurality of music and impression values quantifying impressions of the music; and a music database using the predicted impression value as a key. A music search device comprising: an impression value search unit for searching. 36. An audio signal input unit for inputting an audio signal of a music, a characteristic amount extracting unit for extracting a predetermined characteristic amount from the audio signal, and an impression of the music is quantified from the characteristic amount. 36. The music search device according to claim 35, further comprising: an impression value calculation unit configured to calculate an impression value, wherein the music database stores the acoustic signal and the impression value. 37. The search requirement input means has a music impression value calculation means for calculating an impression value of the input music, and the music impression value calculation means inputs an audio signal of a music to be searched. An audio signal input unit; a characteristic amount extraction unit configured to extract a predetermined characteristic amount from the audio signal; and an impression value calculation unit configured to calculate an impression value of a musical piece from the characteristic amount. 37. The music search device according to claim 35, wherein the music database is searched using the impression value calculated by the value calculation means as a key. 38. The music database stores a feature quantity extracted by the feature extraction means, the search requirement input means inputs a search requirement related to a feature quantity of a search target music, and the impression value search means 37. The music search device according to claim 36, wherein the music database is searched using a requirement regarding a feature amount of the music to be searched as a search key. 39. The method according to claim 39, wherein the impression value calculating means calculates an impression value of the music based on a feature amount of the plurality of music and a coefficient calculated by a multivariate analysis from the impression value of the music. 37. The music search means according to claim 36. 40. The feature amount extracting means calculates a quantity related to a sound generation cycle and a sound intensity included in an audio signal of a music piece as a feature amount.
A music search device according to any one of the preceding claims. 41. The music search according to claim 36, wherein said characteristic amount extraction means calculates a quantity representing a degree of change in the spectrum of an audio signal of the music as a characteristic amount. apparatus. 42. The music search apparatus according to claim 36, wherein the feature quantity extraction means calculates a quantity representing a sounding frequency of an audio signal of the music piece as a feature quantity. . 43. The method according to claim 36, wherein the feature quantity extracting means calculates a quantity expressing a degree of the sound periodicity of the sound signal of the music as a feature quantity. Music search device. 44. The music search device according to claim 36, wherein said feature amount extraction means calculates a quantity representing a beat cycle of the music as a feature amount. 45. The music search apparatus according to claim 36, wherein said feature amount extraction means calculates a quantity expressing a beat cycle ratio of the music as a feature amount. 46. The music search apparatus according to claim 36, wherein the feature amount extraction means calculates a quantity representing the beat intensity of the music as a feature amount. 47. The music search device according to claim 36, wherein said feature amount extraction means calculates a quantity expressing a ratio of beat intensities of the music as a feature amount. 48. The feature amount extracting unit divides an audio signal of a music into a plurality of bands, detects a rising component in each band, and calculates a feature amount using the rising component in each band. The music search device according to any one of claims 36 to 39. 49. The predicted impression value calculating unit calculates a predicted impression value based on a predetermined quantity according to a subjective requirement input by the search requirement input unit. 50. The music search device according to any one of items 48 to 48. 50. The impression value search means calculates a Euclidean distance between the predicted impression value and an impression value of each music stored in the music database, and selects candidate music in order from the music having the smallest Euclidean distance. The music search device according to any one of claims 35 to 48, wherein the music search device outputs the result. 51. The impression value search means calculates a Euclidean distance between the predicted impression value and an impression value of each music stored in the music database, and the Euclidean distance is equal to or less than a predetermined distance. 50. A music composition according to claim 35, wherein the music compositions are output as candidate music compositions in order from the music composition having the smallest Euclidean distance.
A music search device according to the item. 52. The impression value search means calculates a similarity between the predicted impression value and an impression value of a candidate song based on the Euclidean distance, and outputs the similarity together with the candidate song. The music search device according to claim 50 or 51. 53. A music selection means for extracting information on the candidate music output from the impression value search means from the music database and presenting it to a user, and selecting a music from the candidate music, and a music selection by the music selection means. 53. A music reproducing means for extracting a sound signal of the reproduced music from the database and reproducing a part or all of the music so that a user can listen to the music. A music search device according to the item. 54. The music search device according to claim 53, wherein said music selection means selects a music according to a predetermined music selection device. 55. The search requirement input means inputs requirements relating to bibliographic information relating to music such as a song title, the music database stores bibliographic information relating to music, and the impression value searching means uses the bibliographic information as a key. 4. The music database is searched for as
55. The music search device according to any one of items 5 to 54. 56. The music search device according to claim 55, further comprising bibliographic information input means for inputting bibliographic information on music such as a music title, and wherein said music database stores said bibliographic information. 57. The impression value search means, wherein the impression value of the song corresponding to the bibliographic information input by the search requirement input means is used as a search key as a predicted impression value. Music search device as described. 58. The search requirement input means inputs a search device for a requirement relating to bibliographic information such as a song title and a requirement relating to subjectivity such as an image of a song, and the impression value retrieval means searches for bibliographic information according to the retrieval device. 58. The method according to claim 55, wherein the search is performed in combination with a search for a predicted impression value.
A music search device according to the item. 59. An audio signal input unit for inputting an audio signal of a music, a characteristic amount extraction unit for extracting a predetermined characteristic amount from an audio signal output from the audio signal input unit, A music search data registration device, comprising: an impression value calculation unit configured to calculate an impression value obtained by quantifying a music impression. 60. The music search data registration device according to claim 59, wherein said characteristic amount extracting means calculates a quantity related to a sound generation period and a sound intensity included in an audio signal of the music as a characteristic amount. 61. The music search data registration device according to claim 59, wherein said feature amount extraction means calculates a quantity representing a degree of change in the spectrum of an audio signal of the music as a feature amount. 62. The music search data registration device according to claim 59, wherein said feature amount extraction means calculates a quantity expressing a sounding frequency of an audio signal of the music as a feature amount. 63. The music search data registration device according to claim 59, wherein said feature amount extraction means calculates a quantity representing a degree of sound periodicity of an audio signal of the music as a feature amount. 64. The music search data registration device according to claim 59, wherein said feature amount extraction means calculates a quantity representing a beat cycle of the music as a feature amount. 65. The music search data registration device according to claim 59, wherein said feature amount extraction means calculates a quantity expressing a beat cycle ratio of the music as a feature amount. 66. The music search data registration device according to claim 59, wherein said feature amount extraction means calculates a quantity representing the beat intensity of the music as a feature amount. 67. The music search data registration device according to claim 59, wherein said feature amount extraction means calculates a quantity expressing a ratio of beat intensities of the music as a feature amount. 68. The feature amount extracting unit divides an audio signal of a musical piece into a plurality of bands, detects a rising component of each band, and calculates a feature amount using the rising component of each band. 60. The music search data registration device according to claim 59. 69. A computer comprising: a search requirement input means for inputting a user's subjective requirement for a music desired to be searched for and digitizing and outputting the numerical value; A predicted impression value calculating means for calculating a predicted impression value by quantifying the impression; a music database storing sound signals of a plurality of songs and impression values quantifying the impressions of the songs; and Impression value search means for searching the music database, and music search program to function as an impression value. 70. A computer, comprising: an audio signal input unit for inputting an audio signal of a musical composition; a characteristic amount extracting unit for extracting a predetermined characteristic amount from the audio signal; and an impression of the musical composition from the characteristic amount. 70. The music search program according to claim 69, wherein the music search database is made to function as impression value calculation means for calculating a quantified impression value, and wherein the music database stores the acoustic signal and the impression value. 71. A computer, wherein the search requirement input means has a music impression value calculation means for calculating an impression value of the input music, and wherein the music impression value calculation means has an audio signal of a music to be searched. Sound signal input means for inputting, from the sound signal, a feature amount extracting means for extracting a predetermined feature amount, and an impression value calculating means for calculating an impression value of a song from the feature amount, 71. The music search program according to claim 69, wherein the music database is searched using the impression value calculated by the music impression value calculation means as a key. 72. The computer, wherein the music database stores the feature quantity extracted by the feature extracting means, the search requirement input means inputs a search requirement relating to a feature quantity of the music to be searched, and the impression value 71. The music search program according to claim 70, wherein the search means is operable to search the music database using a requirement relating to a feature amount of the music to be searched as a search key. 73. A computer, wherein the impression value calculating means calculates an impression value of the music based on a feature amount of the plurality of music and a coefficient calculated by a multivariate analysis from the impression value of the music. 71. The music search program according to claim 70, wherein the music search program is caused to function. 74. The computer according to claim 70, wherein the feature value extracting means calculates a quantity related to a sound generation cycle and a sound intensity included in an audio signal of the music as a feature value.
1. The music search program according to 1. 75. The computer according to claim 70, wherein the computer is made to function so that the feature quantity extracting means calculates a quantity representing a degree of change in the spectrum of the audio signal of the music as a feature quantity. A music search program according to claim 1. 76. The computer according to claim 70, wherein the computer is made to function so that the feature quantity extracting means calculates a quantity representing a sounding frequency of an audio signal of the music as a feature quantity. Music search program described in section. 77. The computer-readable recording medium according to claim 70, wherein the computer is configured to function so that the feature amount extracting unit calculates a quantity representing the degree of the sound periodicity of the sound signal of the music as the feature amount. A music search program according to any one of the preceding claims. 78. The computer according to any one of claims 70 to 73, wherein the computer is made to function so that the characteristic amount extracting means calculates a quantity representing a beat cycle of the music as a characteristic amount. Music search program. 79. The computer according to any one of claims 70 to 73, wherein the computer is made to function so that the feature quantity extracting means calculates a quantity representing a beat cycle ratio of the music as a feature quantity. Music search program described. 80. The computer according to claim 70, wherein the computer is caused to function such that the feature amount extracting means calculates a quantity representing the beat intensity of the music as a feature amount. Music search program. 81. The computer according to claim 70, wherein the computer is made to function so that the feature quantity extracting means calculates a quantity expressing a ratio of beat intensities of the music as a feature quantity. Music search program described in. 82. The computer, wherein the feature amount extracting unit divides the audio signal of the music into a plurality of bands, detects a rising component of each band, and calculates a feature amount using the rising component of each band. The music search program according to any one of claims 70 to 73, wherein the music search program is caused to function as follows. 83. A computer, wherein the predicted impression value calculating means is operable to calculate a predicted impression value based on a predetermined quantity according to a subjective requirement input by the search requirement input means. The music search program according to any one of claims 69 to 82, characterized by: 84. The computer, wherein the impression value searching means calculates a Euclidean distance between the predicted impression value and an impression value of each music stored in the music database, and calculates a Euclidean distance from the music having a small Euclidean distance. 83. The music search program according to any one of claims 69 to 82, wherein the music search program is made to function so as to output the candidate music in order. 85. A computer, wherein the impression value search means calculates a Euclidean distance between the predicted impression value and an impression value of each music stored in the music database, wherein the Euclidean distance is predetermined. 84. A music piece having a distance less than or equal to the distance is output as a candidate music piece in order from the music piece having the smallest Euclidean distance.
Music search program described in section. 86. The computer, wherein the impression value search means calculates a similarity between the predicted impression value and the impression value of the candidate song based on the Euclidean distance, and outputs the similarity together with the candidate song. A music search program according to claim 84 or 85, wherein the music search program is caused to function. 87. A computer which extracts information on the candidate music output from the impression value search means from the music database, presents the extracted information to a user, and selects a music piece from the candidate music pieces. 89. The music reproducing device according to claim 69, wherein the sound signal of the music selected by the means is extracted from the database, and the music is played as music reproducing means for reproducing a part or all of the music so that the user can listen to the music. A music search program according to any one of the preceding claims. 88. The music search program according to claim 87, wherein the computer causes the music selection means to select a music piece according to a predetermined music selection device. 89. A computer, wherein the search requirement input means inputs requirements relating to bibliographic information relating to music such as a music title, the music database stores bibliographic information relating to music, and the impression value searching means comprises: The music search program according to any one of claims 69 to 88, wherein the music search program is made to function to search the music database using information as a key. 90. The computer according to claim 89, further comprising: bibliographic information input means for inputting bibliographic information relating to a song such as a song title, and wherein the song database functions to store the bibliographic information. Music search program. 91. The computer according to claim 91, wherein said impression value search means functions as a search key with an impression value of the song corresponding to the bibliographic information input by said search requirement input means as a predicted impression value. A music search program according to claim 89 or 90. 92. A computer, wherein the search requirement input means inputs a search device for requirements relating to bibliographic information such as a song name and a requirement relating to subjectivity such as an image of a song, and the impression value search means performs bibliography according to the search device. 92. The music search program according to any one of claims 89 to 91, wherein the music search program is operable to combine a search for information and a search for a predicted impression value. 93. A computer, comprising: an audio signal input unit for inputting an audio signal of a musical composition; a characteristic amount extraction unit for extracting a predetermined characteristic amount from an audio signal output by the audio signal input unit; And an impression value calculation means for calculating an impression value obtained by quantifying the impression of the music, and a music search data registration program for functioning as an impression value calculation means. 94. The music piece according to claim 93, wherein the computer is caused to function such that the feature quantity extraction means calculates a quantity relating to a sounding cycle and a sounding intensity included in an audio signal of the music piece as a feature quantity. Search data registration program. 95. The music search apparatus according to claim 93, wherein the computer is made to function so that the feature quantity extraction means calculates a quantity expressing the degree of change in the spectrum of the audio signal of the music piece as the feature quantity. Data registration program. 96. The music search data according to claim 93, wherein the computer is caused to function such that the feature amount extraction means calculates a quantity expressing the sounding frequency of the sound signal of the music as a feature amount. Registration program. 97. The music piece according to claim 93, wherein the computer is made to function so that the feature quantity extraction means calculates a quantity expressing the degree of the sound periodicity of the sound signal of the music piece as a feature quantity. Search data registration program. 98. The music search data registration program according to claim 93, wherein the computer is caused to function such that the feature amount extracting means calculates a quantity representing a beat cycle of the music as a feature amount. 99. The music search data registration program according to claim 93, wherein the computer is caused to function such that the feature amount extraction means calculates a quantity expressing a beat cycle ratio of the music as a feature amount. . 100. The music search data registration program according to claim 93, wherein the computer is caused to function such that the feature amount extracting means calculates a quantity expressing the beat intensity of the music as a feature amount. 101. The music search data registration according to claim 93, wherein the computer is made to function so that the feature amount extracting means calculates a quantity expressing a ratio of beat intensities of the music as a feature amount. program. 102. A computer, wherein the feature amount extracting means divides an audio signal of the music into a plurality of bands, detects a rising component of each band, and calculates a feature amount using the rising component of each band. 94. The music search data registration program according to claim 93, wherein the program is made to function as follows. 103. A search requirement input means for inputting a user's subjective requirement for a music desired to be searched for, digitizing and outputting the same, and outputting a search target music from the output of the search requirement input means. A predicted impression value calculating means for calculating a predicted impression value by quantifying the impression; a music database storing sound signals of a plurality of songs and impression values quantifying the impressions of the songs; and A recording medium for recording an impression value search means for searching a music database and a music search program for functioning as an impression value search means. 104. A computer, comprising: an audio signal input unit for inputting an audio signal of a musical composition; a characteristic amount extraction unit for extracting a predetermined characteristic amount from an audio signal output by the audio signal input unit; A recording medium for recording a music search data registration program for functioning as an impression value calculating means for calculating an impression value obtained by quantifying the impression of the music.