JP3631650B2 - Music search device, music search method, and computer-readable recording medium recording a music search program - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a music search, and in particular, a part of music similar to a search key from music stored in a database using a part or all of the music input or designated by music input means as a search key. Alternatively, the present invention relates to an entire search apparatus and method.
[0002]
[Prior art]
When performing a similar music search, notes are used as a matching processing unit in the conventional method. In this case, a note refers to a continuation of sounds that fall within a certain threshold value with a fluctuation or rate of change in sound pitch. In addition, each note has a note pitch (pitch), a note length (pitch length), a pitch difference (relative pitch) with a note that fluctuates in time (relative pitch), and fluctuates in time. It holds the ratio of the note length to the note (relative note length).
[0003]
FIG. 9 is a diagram for explaining the principle of a conventional music search using DP matching in units of musical notes. First, in the database construction phase, the music to be searched is input (step S101) and converted into note information (step S102). It is stored in the database (step S103). To search for music stored in the database, a music serving as a search key is input (step S111) and converted into note information (step S112). Next, DP (Dynamic Programming) matching is performed on the note information of all the songs in the database and the note information obtained from the search key by the number of songs in the database (step S113), and the most similar song candidate is selected. The search result is output (step S114).
[0004]
As described above, when music similarity search is performed, DP matching is used in the conventional method (references 1 and 2).
[Reference 1] Tomoya Sonoda, Masataka Goto, Yoichi Muraoka: “Song Retrieval System on the WWW”, IEICE Technical Report, SP97-103, 1998.
[Reference 2] Tetsuya Hatakeyama and Yotsunori Takashima: “Melody search based on humming singing”, IEICE Transactions, Vol. J77-D-II No. 8 pp. 1543-1551, 1994.
In DP matching, the correspondence between the search key and some or all of the songs in the database (candidate partial songs) is required, and between the notes of the search key in correspondence and the notes of the candidate partial songs , The similarity is calculated by the cumulative difference of the note information. In DP matching, a candidate partial song that maximizes the similarity with the search key is searched by dynamic programming.
[0005]
[Problems to be solved by the invention]
In the conventional search method using the note as a unit of matching, the amount of note number errors that can be tolerated is generally limited.
[0006]
Further, in the DP matching of the conventional method, in order to suppress the calculation amount, there is a restriction on the difference in the number of notes between the note string of the search key and the note string of the music in the database that is the similarity calculation target.
[0007]
Therefore, when searching for music, when entering search keys, if you insert or delete more than a certain amount of notes in the search target database, the search accuracy will be extremely poor. was there.
[0008]
Even if a note is not inserted or deleted when a search key is input, there is a high possibility that a note will be inserted or deleted when extracting note information from the search key. There was a problem.
[0009]
In DP matching, the index cannot be used, and all the songs in the database must be searched. Therefore, the amount of error in the number of notes that can be tolerated is limited to reduce the amount of calculation. Even so, there was still a problem that the search took time.
[0010]
On the other hand, in the case of searching using music pieces, the index can be used, so that the search time can be greatly reduced.
[0011]
FIG. 10 is a diagram for explaining the principle of music search using music pieces. This technique is proposed in Japanese Patent Application No. 10-341754 (music search device, music search method, and recording medium on which a music search program is recorded) and has been proposed by the present inventors.
[0012]
First, in the construction of the database, the music to be searched is input (step S201), the input music is converted into note information (step S202), and the note information is divided into a plurality of music pieces (step S203). An index is generated so that a high-speed search can be performed (step S204), and information on the music piece is stored in the database together with the index (step S205).
[0013]
Next, in the music search, a music serving as a search key is input (step S211), the input music is converted into note information (step S212), and the note information is divided into a plurality of music pieces (step S213). Using each of the music pieces as a search key, a similar one is searched from the music pieces stored in the database using an index (step S214). Similarity between some or all of the songs in the database containing the detected music piece by examining the relationship between the music pieces detected as search results and the relationship between the music pieces that became the search key The degree is checked (step S215), and a part or all of the music in the database having a high similarity with part or all of the music that has become the search key is output as a search result (step S216).
[0014]
When searching for music pieces, the number of notes between music pieces for which similarity is calculated is the same, so there is no need to find the correspondence between notes using DP matching. Search was possible.
[0015]
However, when searching for music pieces, there was a limitation that no number of notes was inserted or deleted by the length of the music piece because there was no mistake in the number of notes. That is, there is a problem that the search accuracy is poor if there is no section in which there is no insertion / deletion of the note by the length of the music piece.
[0016]
Further, as a method that does not necessarily require note information, there is a method of performing similarity search after conversion to music information.
[0017]
FIG. 11 is a diagram for explaining the principle of music search without using note information. This method is disclosed in Japanese Patent Application No. 10-329131 (music information retrieval device, music information storage device, music information retrieval method, music information storage method, and recording medium on which these programs are recorded). It is what.
[0018]
First, in the construction of a database, music is input (step S301), converted into music information that can extract musical features (features) from the input music (step S302), and stored in the database. (Step S303).
[0019]
In searching for music, a music serving as a search key is input (step S311), the input music is converted into equivalent music information (step S312), and the music in the database is converted based on the converted music information. A similar search is performed (step S313). Here, as a similarity search, search is performed by comparing the feature quantity of the music of the search key with the feature quantity of the music stored in the database. Calculate the degree. Candidate music pieces including similar ones are output as search results (step S314).
[0020]
However, such a method using music information has a problem that the search accuracy is poor when the tempo of the search key and the song in the database are different.
[0021]
The present invention has been made in view of the above points, and by standardizing the tempo of the music and search key in the database, the search is not performed in units of notes but in units of standardized time. It is an object of the present invention to provide a music searching apparatus and method that are not affected by an error in the number of notes. In the present invention, since it is not necessary to use DP matching, it is possible to perform a search using an index, and a high-speed search is possible. Further, since the tempo is standardized before the search is performed, it is possible to obtain a high search accuracy even when the tempo of the search key and the song in the database are different.
[0022]
[Means for Solving the Problems]
FIG. 1 and FIG. 2 are diagrams for explaining the outline of music similarity search according to the present invention, and FIG. 3 is a principle configuration diagram of the present invention.
[0023]
In the music search, the music information input from the database or the music information converted from the music is standardized and stored in the database and converted from the music or music input as the search key. The main feature is to perform a similar music search after standardizing the tempo of music information.
[0024]
First, in the construction of a database in a search that does not convert to music information, music is input using the music input means 1 (step S10). The input music is tempo normalized by the tempo normalizing means 2 (step S11).
[0025]
Further, in the construction of a database for the search for conversion to music information, after the music is input using the music input means 1 (step S30), it is converted into music information by the music information conversion means 31 (step S31). Thereafter, the tempo standardization means 2 standardizes the tempo (step S32).
[0026]
Tempo standardized music or tempo standardized music information is stored as a database in the music information storage unit 42 by the music storage means 4 (steps S12 and S33).
[0027]
The tempo standardized music or the tempo standardized music information is converted into music information by the music information converting means 31 ′ and stored in the music pieces by the music piece generating means 32 before being stored by the music storage means 4. It may be accumulated after the division process. Further, the index generation means 44 may create an index for speeding up the search for the stored music, music information, and music pieces.
[0028]
The music storage means 4 can store the music before the tempo standardization in the music storage unit 41, and can output the music before the tempo standardization as a search result. .
[0029]
In searching for music, first, a part or the whole of music is input or designated using the music input means 1 (steps S20 and S40). The input music is tempo normalized by the tempo normalizing means 2 (step S21) or converted into music information by the music information converting means 31 (step S41) and then tempo normalized (step S42). . The tempo standardized search key may be further converted into music information by the music information converting means 31 ′ and divided into music pieces by the music piece generating means 32.
[0030]
The search key having undergone the above processing is sent to the music search means 5, and the music search means 5 searches for a part or the whole of the music similar to the search key from the music stored in the database (steps S22 and S43). . If an index is generated in the database, search is performed using the index.
[0031]
In the search using music pieces, the music piece search means 51 searches for similar ones from the music pieces stored in the database using at least one of the music pieces sent as search keys, and outputs the music pieces. 6, the search result is output (steps S23 and S44). When searching using a plurality of pieces of music sent as search keys, after searching for pieces of music in the database similar to each piece of music of the search key, the pieces of music are matched by the piece of music piece matching means 52. It is possible to check the consistency of the search result and use it as the final search result.
[0032]
For the matching of music pieces, the technique described in the specification of the related patent application (Japanese Patent Application No. 10-341754) is adopted. Specifically, the relationship between the music pieces detected as the search results and the relationship between the music pieces that are the search keys of the detected music pieces are examined, and the musical notes before and after the detected music piece are also checked. And the music in the database that contains the detected piece of music and part or all of the piece of music that became the search key by examining the relationship between the musical piece that was the search key for the detected piece of music Investigate the degree of similarity between some or all of.
[0033]
The music input unit 1 may include a rhythm signal transmission unit 11 that emits a signal at a constant time interval or a time interval according to a predetermined rule. In particular, tempo standardization may not be performed with sufficient accuracy for music input as a search key. Therefore, a signal is transmitted by the rhythm signal transmission means 11 at a constant time interval or a time interval according to a predetermined rule, and a music input person inputs music according to the signal, and the input music, By inputting the signal transmitted from the signal transmission unit 11 to the tempo standardization unit 2 at the same time, it is possible to perform accurate tempo standardization.
[0034]
Alternatively, the music input means 1 may have a rhythm signal receiving means 12. This is means for enabling the music input person to receive a rhythm signal that is emitted in accordance with the performance (singing) of the music. The input music and the signal received by the rhythm signal receiving means 12 are simultaneously used for tempo normalization means 2. It is possible to perform an accurate tempo standardization by inputting to.
[0035]
The tempo standardization means 2 is received by the rhythm signal receiving means 12, the music input by the music input means 1, the music information converted by the music information converting means 31, or the signal transmitted by the rhythm signal transmitting means 11. Tempo extraction means 21 for extracting the tempo from the received signal may be provided. In this case, the tempo standardization means 2 performs tempo standardization using the tempo extracted by the tempo extraction means 21.
[0036]
Hereinafter, the operation of the present invention will be described. FIG. 4 is a diagram for explaining tempo standardization. In music, the time length of musical “beat” is defined by the tempo. As shown in the “original music” in FIG. 4A, each music has a tempo, and this tempo may or may not match between the music. In addition, even the same music may differ depending on the performance. Even if the music is the same, it may differ depending on the part of the music.
[0037]
Standardizing the tempo means that, as shown in Fig. 4 (B), in all parts of all performances of all songs, the length of the time is a certain length of time. It is to convert the music.
[0038]
FIG. 5 is a diagram for explaining a unit of matching. When tempo standardization is not performed, musical notes have mainly been the unit of matching. However, as shown in FIG. 5A, when notes are inserted / deleted, there is a problem that the correspondence between the notes to be compared is shifted. DP matching is a method for correcting a shift in correspondence between notes to be compared when notes are inserted / deleted.
[0039]
When a search is actually performed using DP matching, it is necessary to limit the amount of note insertion / deletion allowed, and therefore note insertion / deletion greatly affects the search accuracy. In addition, in the search using DP matching, it is impossible to create an index for performing the search at high speed, so that a long search time is required.
[0040]
On the other hand, when tempo standardization is performed, matching in units of time is possible as shown in FIG. In this case, even if a note is inserted or deleted, there is no deviation in the correspondence relationship, which does not affect the search accuracy. Therefore, it is possible to search with high accuracy. It is also possible to search at high speed by dividing a tempo-standardized music piece into musical pieces and creating an index.
[0041]
As techniques related to the present invention, there are inventions disclosed in Japanese Patent Application Nos. 10-329131 and 10-341754. The present invention is mainly different from the previous invention in that the tempo standardization of the music input to the database is performed when the database is constructed, and the tempo standardization is performed also when the music serving as a search key is input. In particular, the present invention is mainly different from the invention disclosed in Japanese Patent Application No. 10-341754 in that a search is performed in units of time without performing a search in units of musical notes.
[0042]
DETAILED DESCRIPTION OF THE INVENTION
[First Embodiment]
A first embodiment of the present invention will be described. MIDI data is used as music for input to the database. At present, many songs such as karaoke data are available as MIDI data.
[0043]
In MIDI data, note information is described with reference to time signature. And 1 beat Is described separately from the note information as tempo information. Therefore, as standardized time, 1 beat length If the time based on is adopted, tempo standardization of music in the database is realized.
[0044]
In storing in the database, the original MIDI data is first divided into music pieces. As a division method, the length of music pieces is all the same on the standardized time axis, and as shown in FIG. 6 (A), music pieces starting from the beginning of all notes are There is a way to make it generated. Alternatively, as shown in FIG. 6B, there is a method of dividing the musical piece so that it starts at a constant time interval.
[0045]
Also, after dividing into music pieces, a pitch transition vector is generated as music information (features). As shown in FIG. 7, this is a vector in which the standardized time axis is inscribed at equal intervals and the pitches of the sounds being played at the positions of each division are arranged. For the processing by the computer, the pitch is replaced with a numerical value representing the pitch used in MIDI, a frequency, or the like, for example.
[0046]
As the music information (feature amount), in addition to the above, for example, the following can be generated.
[0047]
(1) Pitch transition vector A _i = (A _i1 , A _i2 , ..., a _in ), The average value m of vector elements _i = (A _i1 + A _i2 + ... + a _in ) / N is calculated, and a vector A whose elements are values obtained by subtracting the average value from each element _i '= (A _i1 -M _i , A _i2 -M _i , ..., a _in -M _i ) As a feature value. This vector is expected to be similar even if the key of the search key is different from the key in the database.
[0048]
(2) Pitch transition vector A _i = (A _i1 , A _i2 , ..., a _in ) For the first element a _i1 And a vector A whose elements are values obtained by subtracting this reference value from each element _i '' = (0, a _i2 -A _i1 , ..., a _in -A _i1 ) As a feature value. This vector is expected to be similar even if the key of the search key is different from the key in the database.
[0049]
(3) Pitch transition vector A _i = (A _i1 , A _i2 , ..., a _in ) For each element, the vector A with the value obtained by subtracting the value of the previous element _i '''= (A _i1 , A _i2 -A _i1 , ..., a _in -A _in-1 ) As a feature value. Even if the key of the search key song changes little by little, it is expected that this vector will be similar to the song in the database.
[0050]
After these music information is stored in the database, an index is generated to speed up the search. R-tree or the like is used as an index generation method. In addition, the following are references for R-tree.
[Reference 3] A. Gutman, "R-trees: a dynamic index structures for spatial searching", Proceedings of the ACM SIGMOD International Conference on the Management. 47-57, Boston, 1984.
As for the input of music as a search key, there are methods such as input by humming, input by playing a musical instrument, and input by playing a MIDI keyboard.
[0051]
When inputting music, an electronic metronome is used as a rhythm signal transmitter. This electronic metronome can emit sound, light, characters and figures on a computer display at a specified tempo, and the music input person can input music in accordance with the signal generated. It is like that. This electronic metronome not only emits signals at regular time intervals, but, for example, the first beat of a measure is the tone / volume, light color / light intensity, character type / figure on the display The type can be generated differently from other beats, or a rhythm signal can be generated according to the type of music to be input, such as “samba rhythm”. The tempo at which this signal is generated can be freely set to be slow or fast so that the music input person can easily input the music.
[0052]
Alternatively, a rhythm signal receiving device may be used when inputting music. This means that the music input person can detect movements, signals, etc. consciously / unconsciously when inputting music, such as clapping, stepping, finger up / down movement, body shaking, mouse clicking, keyboard pressing, etc. This is a device
[0053]
Using the input music and the signal obtained from the rhythm signal transmission / reception device, the tempo of the music as a search key is standardized.
[0054]
The division into music pieces and the conversion into music information are the same as the case of music input to the database for music input from the MIDI keyboard.
[0055]
In the input by humming and the input by playing a musical instrument, the music input means 1 inputs music as PCM format data. Then, pitch information is extracted from this data. The pitch information is extracted in accordance with a conventional method for extracting note information from PCM format data.
[Reference 4] Niihara Takamizu, Masakazu Imai, Seiji Iguchi: “Automatic Musical Singing”, Transactions of the Society of Instrument and Control Engineers, Vol. 20 No. 10 pp. [Reference 5] Masanori Mizuno, Masaki Fujimoto, Yotsunori Takashima, Shichiro Tsuruta: "Personal Computer Music System-Automatic Recording of Singing Voices", Information Processing Society of Japan 35th Annual Conference, 5Ff-5 1987.
That is, first, frequency analysis is performed using FFT or autocorrelation processing, peak extraction, power measurement of each peak, and the like are performed. In addition, noise is removed, pitch drift is corrected, and harmonic components are absorbed into the fundamental component, and which sound is played at what point in time on the scale on the time axis. The method of dividing into pieces of music and generating music information after the pitch information is extracted is the same as in the case of inputting music into the database.
[0056]
Note that the music information generation process may be divided into two stages, and tempo standardization may be performed in the middle. For example, after pitch information is first extracted, the tempo is normalized, and then a feature quantity such as a pitch vector is generated.
[0057]
For the search, first, a music piece similar to each piece of music generated from the music input as the search key is searched from the database (music piece search). At that time, the distance between the music piece of the search key and the music piece in the database is obtained for each feature quantity such as a pitch vector. This distance is obtained by setting an appropriate distance such as the Euclidean distance or the Manhattan distance for each feature quantity. In order to speed up the search, an index is used for the search.
[0058]
The distance between music pieces (dissimilarity) is expressed as a value obtained by some method by adding an appropriate weight to the distance of each feature amount. Details are as follows. A set of feature quantities used is FSPACE. If the distance about the feature quantity i between the music piece K and the music piece X is di (K, X) and the weight is Wi, the total distance D (K, X) between the music piece K and the music piece X is The method is to use the following function F.
[0059]
D (X, K) = F ({Wi}, {di (K, X)})
(However, i ∈ FSPACE)
As a specific example of this function F, if a linear sum function is used,
D (X, K) = sum (Wi × di (K, X))
(Where i∈FSSPACE, sum takes the sum over i∈FSSPACE).
[0060]
For example, a function having the following minimum value can be used instead of the linear sum as the function F.
[0061]
D (X, K) = min (Wi × di (K, X))
(Where i∈FSSPACE, min takes the minimum value for i∈FSSPACE)
A mixture of the linear sum and the minimum value can also be used. That is, the feature quantity (FSSPACE) is divided into two types (FSSPACE1, FSPACE2),
FSPACE = FSSPACE1∪FSSPACE2
If
D (X, K) = sum (Wi × di (K, X), min (Wj × dj (K, X))
The total distance can also be obtained using a function such as Here, i∈FSSPACE1, j∈FSSPACE2.
[0062]
Here, as examples of how to calculate the distance between music pieces (dissimilarity), a linear sum is used, a minimum value is used, and a mixture of linear sum and minimum value is used. There are many ways to find the total distance. Of the music pieces in the database, those that are close to the music piece of the search key, that is, those having a high degree of similarity are the results of music piece search.
[0063]
When multiple pieces of music are generated from the search key, the consistency between the pieces of music is checked for the result of the piece search for each piece of music of the search key, and the similarity is calculated when multiple pieces of music are combined. The final search result can be obtained.
[0064]
Now, the music piece K of the search key K _p , K _q It is assumed that music pieces X and Y in the database are similar to (p <q) (K _p ~ X, K _q ~ Y). Music piece X, Y is the search piece music piece K _p , K _q Consistency with respect to satisfies the following conditions.
[0065]
1) Music pieces X and Y must belong to the same song. In the following, this candidate song is A, and music pieces X and Y are A. _r , A _s far.
[0066]
2) Music piece A _r , A _s The temporal order of is the music piece (K _p , K _q ) In the same time order (ie, r <s).
[0067]
3) Music piece A _r , A _s The start time in the normalized time of _Ar , T _As And the music piece K of the search key _p , K _q The start time in the normalized time of _Kp , T _Kq If the start time intervals coincide, that is, T _As -T _Ar = T _Kq -T _Kp It is.
[0068]
In the output of the search result, after obtaining the similarity between the search key and a part or the whole of the music in the database as a candidate for the search result, a part or the whole of the music is ordered in descending order of similarity. Present. The title and musician name of the music can be displayed from the information in the database, and a part or the whole of the music similar to the search key can be played.
[0069]
In the present embodiment, MIDI data can be used as music input to the database. MIDI data is a standardized format, and a great number of music pieces exist as MIDI data, and it is not necessary to perform music pieces for the construction of the database, and the construction of the database is easy. Since MIDI data describes note information based on the time signature, tempo standardization and pitch information extraction can be performed accurately, so it is suitable for use in a music search database.
[0070]
In addition, when inputting music as a search key, the tempo standardization of music as a search key can be accurately performed by using a rhythm signal transmitting / receiving device as an auxiliary.
[0071]
As a special effect of the present embodiment, it is possible to easily and accurately construct a database, and to accurately standardize the tempo of a music used as a search key.
[0072]
[Second Embodiment]
In the second embodiment of the present invention, MIDI data is not used as the music input to the database in the first embodiment, but for example, music CD (compact disc) data, WAVE, A PCM-encoded music such as AIFF or AU, or another encoding format that has been converted to a PCM encoding format is used.
[0073]
In general, in music, sounds are played in time. Since the sound pressure changes greatly at the moment the performance is started, it is possible to extract tempo information by examining the variation of the sound pressure in the music.
[0074]
More specific tempo information extraction methods include the following methods.
[0075]
(1) A method of searching for a volume level peak with a small sampling frequency such as 100 Hz or 10 Hz and extracting the tempo from the peak interval. If the sampling frequency of the original PCM data is 44100 Hz, for example, a method is considered in which data obtained by adding 1000 samples is created and the peak of the volume level is searched.
[0076]
(2) A method of examining the transition of the change rate of the volume level at the small sampling frequency, searching for a portion where the change rate increases rapidly, and extracting the tempo from the interval of the rapidly increasing portion.
[0077]
(3) A method of searching for a peak by performing frequency analysis on the data of the small sampling frequency in a long section such as 10 seconds. For example, if a peak appears at 2 Hz, the length of one beat of this music is 0.5 seconds.
[0078]
(4) A method of extracting tempo information by selectively extracting the sounds of musical instruments such as drums that engrave rhythms and analyzing fluctuations in the sound pressure of these sounds.
[0079]
As a method for generating a feature value such as pitch information from the input music, the pitch information and the like are extracted from the music input by humming or playing a musical instrument in the first embodiment. However, unlike in the case of humming or playing a single instrument, in music CD data, etc., multiple notes are emitted at the same time, so the pitch to be searched (the pitch of the sound corresponding to the melody) Only need to extract. Therefore, after performing FFT and autocorrelation processing, a method to extract the strongest peak of sound pressure, or a band filter that extracts only the peak of the frequency band of a musical instrument or singing voice that plays melody or rhythm There is a way to use.
[0080]
The effects of the present embodiment are as follows. By using music encoded by the PCM method, which is a more general music encoding method than MIDI, as the music to be input to the database, the amount of music included in the database is made more substantial. Is possible. Further, not only commercially available music CDs, but also songs that are personally played with musical instruments can be included in the database.
[0081]
[Third Embodiment]
In the third embodiment of the present invention, when the music in the database and the music of the search key are divided into pieces of music in the first and second embodiments described above, Instead of generating a transition vector, a pitch distribution in the target musical piece is obtained, and a feature vector is generated based on this distribution.
[0082]
When the pitch of a musical piece changes as shown in FIG. 8A, the unit of normalized time is converted into a number once, and the pitch distribution diagram as shown in FIG. And pitch distribution vectors can be created by arranging the frequencies of pitches (or pitch ranges). This can be searched as music information (features).
[0083]
As the music information (feature amount), in addition to the above, for example, the following can be generated.
[0084]
(1) Pitch transition vector A used in the first embodiment _i A vector obtained by subtracting the average value of elements from each element of _i Distribution of element values for '.
[0085]
(2) Pitch transition vector A used in the first embodiment _i A vector A obtained by subtracting the value of the first element from each element of _i Distribution of element values for ''.
[0086]
(3) Pitch transition vector A used in the first embodiment _i A vector obtained by subtracting the value of the previous element from each element of _i Distribution of element values for '''.
[0087]
In each of the above distributions, the frequency corresponding to 0 corresponding to the horizontal axis in the pitch distribution diagram of FIG. In each of the above distributions, when the value corresponding to the horizontal axis of the pitch distribution diagram of FIG. 8B is limited to a predetermined range and data exceeding the range is deleted, Sometimes added to boundary values.
[0088]
Using these feature vectors, a similarity search is performed after music piece search and music piece matching according to the method of the first embodiment.
[0089]
Regarding the effect of this embodiment, since the number of dimensions of the feature vector can generally be reduced as compared with the case of the first embodiment, the search speed can be improved. In addition, since the time information is lost within the range of the music piece, it is possible to search even if there is some error between the music piece in the database and the start point of the music piece generated from the search key. . On the other hand, since time information is lost, even if it is not similar in the pitch transition vector, it may be similar in the pitch distribution vector. Therefore, it is possible to first perform a search by the method according to the present embodiment, and perform a narrowing search for performing a search according to the first embodiment from the search results.
[0090]
【The invention's effect】
As described above, according to the present invention, the searcher can freely input a part or the whole of the music piece by humming or playing a musical instrument, so that the music piece intended by the searcher can be accurately and quickly retrieved from the database. It becomes possible to search.
[0091]
According to the present invention, for example, in a system such as karaoke or a music-on-demand system (automatic jukebox), even if the user cannot remember the name, composer name, performer name, etc. If you remember even the melody of the song, you can select the desired song, which is extremely effective in this field.
[Brief description of the drawings]
BRIEF DESCRIPTION OF DRAWINGS FIG. 1 is a diagram for explaining the outline of a similar music search (when conversion to music information is not performed) according to the present invention.
FIG. 2 is a diagram for explaining an outline of a similar music search (in the case of conversion to music information) according to the present invention;
FIG. 3 is a principle configuration diagram of the present invention.
FIG. 4 is a diagram for explaining tempo normalization.
FIG. 5 is a diagram for explaining a unit of matching.
FIG. 6 is a diagram illustrating division into music pieces.
FIG. 7 is a diagram illustrating conversion into music information.
FIG. 8 is a diagram for explaining conversion to music information (pitch distribution);
FIG. 9 is an explanatory diagram of the principle of music search (using DP matching) by a conventional method.
FIG. 10 is a diagram for explaining the principle of music search using a music piece related to the present invention.
FIG. 11 is a diagram for explaining the principle of music search without using note information related to the present invention.
[Explanation of symbols]
1 Music input means
11 Rhythm signal transmission means
12 Rhythm signal receiving means
2 Tempo standardization means
21 Tempo extraction means
31 Music information conversion means
31 'Music information conversion means
32 Music piece generating means
4 Music storage means
41 Music storage unit
42 Music information storage
43 Music piece storage
44 Index generation means
5 Music search means
51 Music piece search means
52 Music piece matching means

Claims (14)

In a search device for searching for music,
A music input means for inputting music;
Tempo standard that converts the input music part and normalizes the tempo so that the length of one beat is a certain time length in all parts of all music input by the music input means And
Music storage means for storing at least the music standardized by the tempo normalization means;
Based on a part or the whole of the music that has been standardized by the tempo standardization means from a part or the whole of the music inputted as a search key, a part or the whole of the music stored in the music storage means A music search means for searching for similar music by calculating the distance between feature amounts of temporally corresponding positions of music that are tempo-standardized from each other;
A music search apparatus comprising music output means for outputting music searched by the music search means. The music search device according to claim 1,
Music information converting means for converting the music input by the music input means into music information;
The music search apparatus characterized in that the tempo standardization means performs tempo standardization after the music is converted into music information by the music information conversion means. The music search device according to claim 1 or 2,
The music input means may be a rhythm signal transmission means for transmitting a signal at a constant time interval or a time interval according to a predetermined rule, or a rhythm that allows the music input person to input the rhythm of the music being input together. A music search apparatus comprising signal receiving means. The music search device according to any one of claims 1 to 3,
The tempo standardization means includes a tune input by the tune input means, music information converted by the music information conversion means, a signal transmitted by the rhythm signal transmission means, and a signal received by the rhythm signal reception means. Tempo extraction means for extracting the tempo using at least one of
A music search device characterized by performing tempo standardization using an extracted tempo. The music search device according to any one of claims 1 to 4,
The music search means searches for a part or the whole of the tempo standardized music stored in the music storage means based on a part or the whole of the tempo standardized search key. The degree of similarity between a part or all of the standardized search key and part or all of the tempo standardized music to be searched that has the same length as part or all of the tempo standardized search key The music search device characterized by searching. The music search device according to any one of claims 1 to 4,
The above-mentioned tempo-standardized music or tempo-standardized music information has a length including at least a plurality of notes, and the music pieces have the same length on the standardized time axis. A music piece generating means for dividing,
A music search device characterized in that the music storage means stores information on music pieces divided into music pieces. The music search device according to claim 6,
When the music piece generating means generates a music piece, a music piece for the same musical piece is generated so that the music piece starts at a fixed time interval, or generated so that the music piece starts from the beginning of a note. A music search device characterized by: The music search device according to claim 6 or 7,
The music search means, when performing a search, has a temporally corresponding position between the music piece generated from the music inputted as a search key and the music piece of the music stored in the music storage means. Music pieces similar to each piece of music generated from the input music piece are searched from the music storage means by calculating the distance of the feature amount, and the final search is performed from the relationship between the music pieces as a result. A music search apparatus comprising music piece search means for guiding a result. In a search device for searching for music,
A music input means for inputting music;
Tempo standard that converts the input music part and normalizes the tempo so that the length of one beat is a certain time length in all parts of all music input by the music input means And
Music piece generating means for dividing the tempo-standardized music piece into music pieces having a length including at least a plurality of notes and all having the same length on the standardized time axis; ,
Music storage means for storing music divided into music pieces by the music piece generating means;
The input music piece is calculated by calculating the distance of the characteristic amount of the position corresponding to the time between the music piece generated from the music piece inputted as the search key and the music piece of the music piece stored in the music storage means. Music pieces similar to each of the music pieces generated from the above are searched from the above music storage means, and the final search result is derived from the relationship between the music pieces as a result, and is input as a search key. A music search means for searching for music similar to the music;
A music search apparatus comprising music output means for outputting music searched by the music search means. The music search device according to claim 8 or 9,
The music search means includes music piece matching means for checking the consistency between the relationship between the music pieces of the search key and the relationship between the music pieces of the search result and then leading to a final search result. Music search device. The music search device according to any one of claims 1 to 10,
A music search apparatus comprising music information conversion means for converting into music information before storing in the music storage means. The music search device according to any one of claims 1 to 11,
When storing music, it has an index generation means for generating an index for speeding up the search,
The music search device, wherein the music search means uses an index when searching. A search method for searching for music by a computer comprising at least music input means, tempo normalization means, music storage means, music search means, and music output means,
A process in which the music input means inputs the music;
The tempo normalization means converts the inputted music part so that the length of one beat becomes a certain time length in all parts of all the music inputted by the music input means, The process of standardizing the tempo,
A process in which the music storage means stores music that has been tempo normalized by at least the tempo normalization means;
A part or the whole of the stored music based on a part or the whole of the music that has been standardized by the tempo standardizing means from a part or the whole of the music that is input as a search key by the music search means. Searching for similar music by calculating the distance of the feature quantities of the temporally corresponding positions of music that are tempo-standardized from each other,
A music search method, wherein the music output means includes a step of outputting the music searched by the music search means. A program recording medium recording a program used for searching for music by a computer,
Said computer
A music input means for inputting music;
Tempo standard that converts the input music part and normalizes the tempo so that the length of one beat is a certain time length in all parts of all music input by the music input means And
Music storage means for storing at least the music standardized by the tempo normalization means;
Based on a part or the whole of the music that has been standardized by the tempo standardization means from a part or the whole of the music inputted as a search key, a part or the whole of the music stored in the music storage means A music search means for searching for similar music by calculating the distance between feature amounts of temporally corresponding positions of music that are tempo-standardized from each other;
As a music output means for outputting the music searched by the music search means,
A computer-readable recording medium storing a music search program for functioning.

Images