JP5487718B2 - Sound material search device - Google Patents

Description

  The present invention relates to a technique for supporting search of sound material.

  There is a performance apparatus that has a database that collects sound materials that are fragmentary sound waves, and that edits and plays music using sound materials retrieved from the database. The database of this type of performance device associates a plurality of types of sound materials with a plurality of types of feature values obtained for each of them. Various techniques have been proposed for facilitating the search for sound material based on feature quantities in this type of performance device. For example, in the performance device disclosed in Patent Document 1, each sound material in the database is mapped to a two-dimensional coordinate system in which two types of feature values designated by the user are set as an X axis and a Y axis. A search screen in which symbols indicating materials are arranged in the two-dimensional space is displayed. When the user designates a position corresponding to two desired feature quantities on the search screen with a pointing device, the sound material indicated by each symbol displayed within the predetermined distance of the designated point is selected, and a plurality of performance materials are used. Assigned to each operator.

JP 2008-225200 A “Tenorion”, [online], Yamaha Corporation, Internet Search on April 28, 2009, <URL: http: //www.yamaha.co.jp/design/tenori-on/>

According to the technique disclosed in Patent Document 1, since the characteristics of various sound materials can be grasped as positions in the space, an operation of finding a desired sound material is facilitated. However, when the found sound material is close to the desired one but not the desired one, there is a problem that trial and error of indicating the position corresponding to the feature amount in the search screen must be repeated again. It was.
The present invention has been devised under such a background, and provides a sound material search device capable of efficiently searching for a desired sound material from sound materials stored in a database. With the goal.

  The present invention relates to a storage unit that stores each sound material and a feature vector indicating the characteristics of the sound material in association with each other, an operation unit, a display unit that displays a symbol indicating the sound material, and the operation unit. One or a plurality of sets of feature amounts obtained by interpolating between a plurality of feature amount vectors stored in the storage means in association with the plurality of reference sound materials, using a plurality of sound materials designated by operation as a reference sound material Control means for retrieving one or more sound materials having a feature vector most similar to the vector from the storage means, and displaying a plurality of symbols indicating the sound materials as the search results on the display means. Provided is a sound material retrieval apparatus characterized by:

  According to the present invention, it is possible to search for a sound material having an intermediate characteristic of a plurality of sound materials, so when there are a plurality of sound materials close to a desired one among searched sound materials, A desired sound material can be efficiently searched using these sound materials.

It is a block diagram which shows the structure of the sound material retrieval apparatus which is one Embodiment of this invention. It is a figure which shows the data structure of the sound material database memorize | stored in the hard disk of the same sound material search device. It is a figure which shows the search screen displayed on the display part of the same sound material search device. It is a figure which shows the search screen displayed on the display part of the same sound material search device. It is a figure which shows the search screen displayed on the display part of the same sound material search device. It is a figure which shows the search screen displayed on the display part of the same sound material search device. It is a figure which shows the search screen displayed on the display part of the same sound material search device. It is a figure which shows the search screen displayed on the display part of the same sound material search device. It is a figure which shows the search screen displayed on the display part of the same sound material search device. It is a figure which shows the search screen displayed on the display part of the same sound material search device.

  Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing a configuration of a sound material searching apparatus 10 according to an embodiment of the present invention. The sound material search apparatus 10 is obtained by installing the sound material search program 28 according to the present embodiment in a personal computer. Here, the sound material search program 28 has a function of searching the sound material used for composing the music from the database in accordance with the operation performed by the user, and editing and playing the music using the searched sound material. It is an application program. Note that the sound material in the present embodiment is a sound waveform of a section corresponding to one beat in a song or a sound waveform of a section further divided into one beat.

  As shown in FIG. 1, the sound material search device 10 is connected to a sound system 90 via an interface 11. The operation unit 13 in the sound material searching apparatus 10 includes a mouse 14 and a keyboard 15. The display unit 16 is a liquid crystal display.

  The control unit 20 includes a CPU 21, a RAM 22, a ROM 23, and a hard disk 24. The CPU 21 executes programs stored in the ROM 23 and the hard disk 24 while using the RAM 22 as a work area. The ROM 23 is a read-only memory that stores an IPL (Initial Program Loader) and the like.

The hard disk 24 stores sound material databases 25 and 26, a rhythm pattern database 27, and a sound material search program 28.
The sound material database 25 is a collection of records each corresponding to a sound material (hereinafter referred to as “edge sound”) with a clear attack and a strong edge feeling among the sound materials included in the sound waveform of the music. The sound material database 26 is an aggregate of a plurality of records respectively corresponding to each sound material (hereinafter referred to as “dust sound”) having a strong dusty feeling among the sound materials included in the sound waveform of the music. In addition to these two sound material databases 25 and 26, another database composed of a plurality of records respectively corresponding to sound materials other than edge sound and dust sound may be provided.

FIG. 2 is a diagram showing the data structure of the sound material database 25. As shown in FIG. 2, a record corresponding to one edge sound in the sound material database 25 is obtained by analyzing a sample string of a sound waveform of a section including an edge sound in a sound waveform of a music piece and the sound waveform. In addition, each of the following eight types of feature amounts has nine fields.
a1. High frequency intensity P _high
This is the intensity of the high frequency component contained in the sound material.
b1. Mid-high frequency intensity P _mid-high
This is the intensity of the frequency component in the mid-high range included in the sound material.
c1. Mid-low range strength P _mid-low
This is the intensity of the mid-low frequency component contained in the sound material.
d1. Low frequency intensity P _low
This is the intensity of the low frequency component contained in the sound material.
e1. Dissonance P _disson
This is an index value indicating the degree of harmony of the sound material.
f1. Spectral Complexity P _complex
This is an index value indicating the complexity of the sound material.
g1. Peak position P _peak-posi
This represents the time at which the amplitude of the sound waveform of the sound material reaches its peak as a relative time based on the time of the first sample of the sound material.
h1. Peak intensity P _peak-val
This is the amplitude of the peak in the sound material.

In addition, a record corresponding to one dust sound in the sound material database 26 includes a sample string of a sound waveform in a section including a dust sound in the sound waveform of the music, and the above-described a1. ~ F1. The seven types of feature quantities P _high , P _mid-high , P _mid-low , P _low , P _disson , and P _complex are included.
Hereinafter, a feature quantity vector having the feature quantities P _high , P _mid-high , P _mid-low , P _low , P _disson , P _complex , P _peak-posi , and P _peak-val as elements is referred to as a feature quantity vector P. .
In addition, instead of including the sample sequence of the sound waveform of the sound material in the record indicating the sound material, the information indicating the location of the music including the sound waveform in the hard disk 24 and the section in which the sound waveform appears in the music are shown. Information may be included.

  The rhythm pattern database 27 is a database that stores a plurality of rhythm pattern data. Here, one rhythm pattern data is composed of 16 types of track data Track-k (k = 1 to 16). Each track data Track-k (k = 1 to 16) is a time-series pattern instructing ON (pronunciation) / OFF (non-pronunciation) of a rhythm sound in each track constituting one rhythm pattern.

The sound material search program 28 is a program that causes the CPU 21 to execute the following three processes.
a2. Search Control Processing This is a search screen that is a screen in which symbols SM-i (i = 1, 2,...) That indicate the searched sound material are mapped and the sound material is searched from the sound material databases 25 and 26. It is a process to display on.
b2. Music performance processing This is processing for causing the sound system 90 to sound the sound material indicated by the symbol SM-i designated on the operation unit 13 among the symbols SM-i (i = 1, 2,...) Displayed in the search screen. It is.
c2. Rhythm performance processing In this process, the rhythm pattern data designated by the user through the operation of the operation unit 13 is read from the rhythm pattern database 27 and stored in the RAM 22, and the sound material designated by the user through the operation of the operation unit 13 is stored in each rhythm pattern data. In this process, the rhythm pattern data is automatically reproduced by using the assigned sound material as the rhythm sound and assigned to the track data Track-k (k = 1 to 16).

  One of the features in the present embodiment is that three types of search, that is, a feature amount search, a similarity search, and a gradation search, can be executed in order to facilitate the search for a desired sound material by the user. The feature quantity designation search is a search for finding a sound material having a feature quantity vector P closest to the feature quantity vector P designated by the user. The similarity search is a search for searching for a predetermined number of sound materials having a feature vector P close to the feature vector P of one sound material designated by the user. The gradation search is a search for searching for sound materials having a feature amount vector P intermediate between two sound materials specified by the user by the number specified by the user. In the feature quantity designation search, the feature quantity vector P is directly designated as a search condition, whereas in the similarity search and gradation search, the user directly designates the sound material as the search condition. In this respect, it can be said that the similarity search and the gradation search are different kinds of searches from the feature amount designation search. Details of these three searches will be described later.

  Next, the operation of this embodiment will be described. In the present embodiment, when the user activates the sound material search program 28, the CPU 21 displays a search screen on the display unit 16 in the search control process. As shown in FIG. 3, the search screen has a search operation area 31 and a performance operation area 32. The search operation area 31 is an area indicating a two-dimensional coordinate system composed of an x-axis and a y-axis. Near the upper end of the search operation area 31, 16 frames 5-k (k = 1 to 16) are arranged. In the performance operation area 32, a button 41 for instructing the start of automatic performance of music, an up / down control button 42 for specifying BPM (Beats Per Minute) for determining the tempo of automatic performance, and a rhythm pattern are specified. A pull-down menu 43 and a button 44 for instructing the saving of the music are arranged.

  When the search screen is displayed on the display unit 16, the user searches for sound materials from the sound material databases 25 and 26 and selects 16 sound materials to be used for performance from among them, and 16 materials determination work. Performing music using the sound material of the instrument manually or automatically. Below, operation | movement of this embodiment in a material determination work and a performance work is demonstrated.

(1) Material Determination Work In the material determination work, the user can cause the CPU 21 to execute a feature amount designation search as follows. First, the user selects one symbol SM-i displayed in the search operation area 31 and then displays the operation window 51 in the area 31. When no symbol SM-i is displayed in the search operation area 31, the user can use a feature vector P (for example, feature values P _high , P _mid-high , P, which is preset as a default value). A symbol SM-i indicating a sound material having a feature quantity vector P) in which all of _mid-low , P _low , P _disson , P _complex , P _peak-posi , and P _peak-val are 0 is entered in the search operation area 31. First, the displayed symbol SM-i is selected. As shown in FIG. 4, the operation window 51 includes eight types of feature amounts corresponding to sliders SLDR _high , SLDR _mid-high , _{SLDR mid-low} , _{SLDR low} , _{SLDR disson} , _{SLDR complex} , _{SLDR peak-posi} , and _{SLDR peak. This} is a window in which _-val and a button 52 _labeled "Search" are arranged. The user, each slider SLDR _high of the operating window _{51, SLDR mid-high, SLDR} mid-low, SLDR low, SLDR disson, SLDR complex, SLDR peak-posi, SLDR peak-val desired sound position of the knob After adjusting according to the size of the feature that the material will have, the button 52 in the window 51 is selected.

When this operation is performed, in the search control process, the CPU 21 performs a first process, a second process, and a third process described below.
In the first process, the CPU 21 determines a search condition SC _{char associated} with the symbol SM-i selected by the user. More specifically, the CPU 21 positions the sliders SLDR _high , SLDR _mid-high , _{SLDR mid-low} , _{SLDR low} , _{SLDR disson} , _{SLDR complex} , _{SLDR peak-posi} , SLDR _{peak-val in} the operation window 51. Are converted into feature quantities, and a feature quantity vector P having these feature quantities as elements is set as a search condition SC _char .

In the second process, a sound material record having the shortest Euclidean distance from the feature vector P constituting the search condition SC _char is searched from the sound material databases 25 and 26, and the searched record is a symbol SM selected by the user. Write to the RAM 22 as the latest search result SR of the sound material indicated by -i. Here, as described above, each record in the sound material database 25 includes the feature values P _high , P _mid-high , P _mid-low , P _low , P _disson , P _complex , P _peak-posi , P _peak-val . Therefore, the Euclidean distance from the 8-dimensional feature vector P constituting the search condition SC _char can be obtained, whereas each record in the sound material database 26 has the feature values P _high and P _mid-high. , P _mid-low , P _low , P _disson , and P _complex only, the Euclidean distance from the eight-dimensional feature vector P constituting the search condition SC _char cannot be obtained. Therefore, in the present embodiment, the CPU 21 uses a feature quantity vector P constituting a search condition SC _char (same for search conditions SC _gra and SC _sim described later) for each record in the feature quantity database 26 as a six-dimensional feature quantity. The Euclidean distance between each is determined as a vector P (that is, the feature amounts P _peak-posi and P _peak-val are not used).
In the third process, the CPU 21 outputs the sound material sample sequence that is the latest search result SR to the sound system 90 via the interface 11 and causes the system 90 to generate the sound material.

The user, when a strong sound material (strength is high sound material of the frequency components in the low frequency range) want the bass moves the slider SLDR _low to a position that indicates the larger feature value P _low, strong sound with treble sense If you want material (sound material with high frequency component intensity), move the slider SLDR _high knob to a position that indicates a larger feature P _high, and perform feature-specific search while changing the settings. The instruction is repeated, and it is confirmed whether or not the sound material generated from the sound system 90 is close to a desired one after the feature value designation search. When a sound material that is not a desired sound material but is close to it is found, the user can cause the CPU 21 to perform a similarity search using the sound material as follows.

First, the user selects a symbol SM-i of a sound material that is felt close to a desired sound material, and then presses a key for similarity search on the keyboard 15.
Thereby, the CPU 21 performs first, second, and third processing described below.
In the first process, the CPU 21 determines a search condition SC _{sim associated} with the symbol SM-i selected by the user. More specifically, the CPU 21 sets the feature amount P of the record written in the RAM 22 as the latest search result SR of the sound material indicated by the symbol SM-i as the search condition SC _sim .

In the second process, the CPU 21 searches the sound material databases 25 and 26 for records of the top eight sound materials having the short Euclidean distance from the feature vector P constituting the search condition SC _sim, and the eight of them. The record is written in the RAM 22 as the first to eighth search results SR of the sound material having the feature vector P close to the sound material indicated by the symbol SM-i selected by the user.

  In the third process, as shown in FIG. 5, the CPU 21 has a symbol SM-i (symbol SM-1 in the example of FIG. 5) selected by the similarity search operation near the upper left, the left, the lower left, and the lower. Eight symbols SM-i (in the example of FIG. 5) indicating sound materials that are the first to eighth search results SR at positions near, lower right, right, upper right, and upper, respectively. Symbols SM-2, SM-3, SM-4, SM-5, SM-6, SM-7, SM-8, SM-9) are displayed in order. Here, FIG. 5 shows a search screen when five of the eight symbols SM-i are displayed. In FIG. 5, symbols SM-2, SM-3, SM-4, SM-5 and SM-6 shown as solid line rectangles are already displayed symbols SM-i and shown as chain line rectangles. Symbols SM-7, SM-8, and SM-9 are undisplayed symbols SM-i. The arrow lines indicate the display order (symbol SM-7 → symbol SM-8 → symbol SM-9) of the undisplayed symbols SM-7, SM-8, SM-9.

  The user confirms whether the desired sound is included in the eight sound materials obtained by the similarity search. Specifically, the user designates the selected symbol SM-i and the eight symbols SM-i displayed around it by the mouse 14 one by one. At this time, the CPU 21 reads the sound material record indicated by the symbol SM-i designated by the user with the mouse 14 from the RAM 22 and sends a sound wave sample sequence of the sound material to the sound system 90 to generate the sound material. The user confirms the sound material that is pronounced in this way. The user does not include the desired material among the nine sound materials that are sounded in this way, but there may be a desired material between the two sound materials of the nine sound materials. When it is felt that there is not, gradation search is executed by the CPU 21.

  As shown in FIG. 6, in the gradation search, the user uses two symbols SM-i displayed in the search operation area 31 (symbol SM-2 and symbol SM-6 in the example of FIG. 6). Are moved to positions where an appropriate distance d is left between them, and then each of them is sequentially selected as a gradation search start point and end point, and a gradation search key on the keyboard 15 is pressed.

In the search control process, when the gradation search is instructed, the CPU 21 performs the following first, second, and third processes.
In the first process, the CPU 21 sets a symbol SM-i selected as a starting point by the user (hereinafter referred to as a starting point symbol SM-is) and a symbol SM-i selected as an end point (hereinafter referred to as an end point symbol SM-ie). Based on the above, the search condition SC _gra is determined.

More specifically, the CPU 21 obtains the search number Num by substituting the distance d from the start point symbol SM-is to the end point symbol SM-ie into the following equation (1). In the following equation (1), d _SM is a width corresponding to one symbol _SM .
Num = d / d _SM (1)

Then, the CPU 21 uses the sound material indicated by the start point symbol SM-is and the end point symbol SM-ie as the reference sound material, and the feature amount vector P and the end point symbol of the sound material indicated by the start point symbol SM-is in the feature amount vector space. Num point interpolation is performed between the sound material feature vector P indicated by SM-ie, and each num point vector vector P _int -h (h = 1 to Num: h) obtained by this interpolation. Is an index indicating the rank of the feature vector P with respect to the sound material indicated by the start point symbol SM-is) as search conditions SC _gra .

In the second process, the CPU 21 searches the sound material databases 25 and 26 for a sound material record having a short Euclidean distance from the feature vector VP _int -h (h = 1 to Num) constituting the search condition SC _gra . The searched Num records are written in the RAM 22 as search results SR of Num sound materials having intermediate characteristics between the sound materials indicated by the start point symbol SM-is and the end point symbol SM-ie.

More specifically, the CPU 21 searches the sound material databases 25 and 26 for the record of the sound material having the shortest Euclidean distance from the feature vector P _int −1, and the searched record is the first search result SR. Is written into the RAM 22.

Next, the CPU 21 searches the sound material databases 25 and 26 for a sound material record having the shortest Euclidean distance from the feature vector P _int -2, and the retrieved record is a record that forms the first search result SR. If they are different, the record is written in the RAM 22 as the second search result SR. In addition, when the record of the sound material having the shortest Euclidean distance with the feature amount vector P _int -2 is the same as the record forming the first search result SR, the CPU 21 stores the sound material having the next shortest Euclidean distance. Is re-searched from the sound material databases 25 and 26, and the re-searched record is written in the RAM 22 as the second search result SR.

Thereafter, the CPU 21 repeats the same search as the case of the feature vector P _int -2 with respect to the feature vector P _int −3, P _int −4... P _int −Num, and records obtained by each search are stored in the third record. Are written in the RAM 22 as the .sup.th to Num.sup.th search results SR.

In the third process, as shown in FIG. 7, the CPU 21 starts from the start point symbol SM-is (symbol SM-2 in the example of FIG. 7) to the end point symbol SM-ie (symbol SM-6 in the example of FIG. 7). ) Num symbols SM-i (SM-10, SM-11 in the example of FIG. 7) indicating the sound material which is the search result SR from the first to the Numth in the second process. SM-12, SM-13, SM-14, SM-15) together with is displayed in the order, to sound the sound material is results SR from the first to the Num-th to the sound system 90.

  The user checks whether the desired number is included in the Num sound materials generated from the sound system 90 after the gradation search operation. Specifically, the user instructs each symbol SM-i indicating Num sound materials one by one using the mouse 14. At this time, the CPU 21 reads the sound material record indicated by the symbol SM-i designated by the user with the mouse 14 from the RAM 22 and sends a sound wave sample sequence of the sound material to the sound system 90 to generate the sound material. The user confirms the sound material that is pronounced in this way. If the desired sound material is included in the Num sound materials, the user can select the desired sound SM-i displayed in the search operation area 31 as shown in FIG. The symbol SM-i indicating the sound material (symbol SM-12 in the example of FIG. 8) is selected, and the selected symbol SM-i is selected from any of the frames 5-k (k = 1 to 16) in the region 31 ( In the example of FIG. 8, the frame is moved to the frame 5-5).

Thereby, the CPU 21 associates the sample sequence of the sound material, which is the sound material search result SR indicated by the symbol SM-i moved to the frame 5-k, with the identifier of the frame 5-k, and stores the storage area in the RAM 22. Write to AR. The storage area AR is a storage area reserved for storing a sample sequence of sound materials used in the music performance process and the rhythm performance process.
The user searches for a desired sound material while properly using feature amount designation search, similarity search, and gradation search until all of the frames 5-k (k = 1 to 16) are filled with the symbols SM-i. When all of the frames 5-k (k = 1 to 16) are filled with the symbols SM-i, the material determination work is finished, and the next performance work is started.

(2) Performance work In the performance work, the user causes the CPU 21 to execute manual performance or automatic performance.
In the manual performance, the user is assigned a role as a performance key key-k (k = 1 to 16) corresponding to each of the frames 5-k (k = 1 to 16) among the keys on the keyboard 15. Press the object at the timing you want the sound material to sound.
In the music performance processing, the CPU 21 sends the sound system 90 via the interface 11 to the sample sequence of the sound material stored in the storage area AR in association with the identifier of the frame 5-k each time the performance key key-k is pressed. The sound material is generated from the system 90.

  When causing the CPU 21 to perform automatic performance, the user designates the button 41 with the mouse 14. The user can change the tempo with the up / down control button 42 or change the rhythm pattern data with the pull-down menu 43 before or after instructing the button 41.

  In the rhythm performance processing, the CPU 21 reads out the rhythm pattern data designated by the user through the operation of the operation unit 13 from the rhythm pattern database 27 and stores it in the RAM 22, and the track data Track-k (16-track data constituting the rhythm pattern data). k = 1 to 16) are repeatedly reproduced. At that time, the CPU 21 uses each sound material associated with the frame 5-k (k = 1 to 16) as the rhythm sound of the k-th track (k = 1 to 16). That is, when the track data Track-k of the k-th track among the track data Track-k (k = 1 to 16) read from the RAM 22 indicates ON of the rhythm sound, it is associated with the frame 5-k. The sample sequence of the sound material is read from the storage area AR and sent to the sound system 90 for sound generation.

  The user can also perform an audio file writing operation. In the audio file writing operation, the symbol SM-i in any one or a plurality of frames 5-k in the frame 5-k (k = 1 to 16) is selected, and the button 44 is selected.

  In the music performance process, when an audio file writing operation is performed, the CPU 21 converts a sample string of sound material indicated by the symbol SM-i selected by this operation into an audio file of Waveform Audio Format (hereinafter referred to as “WAV format”). The converted audio file is stored in the hard disk 24. The audio file stored in the hard disk 24 can be used in various ways. For example, the audio file may be copied to the hard disk of another computer in which the sound material search program 28 is installed and used for composition by the computer. Further, an audio file may be copied to a memory in a performance device called Tenorion (registered trademark) proposed by the applicants of the present invention, and the copied audio file may be used for performance of a song by Tenorion. See Non-Patent Document 1 for details of Tenorion.

In the embodiment described above, in the gradation search, the feature amount vector P obtained by interpolating the feature amount vectors P of the sound material indicated by the two symbols SM-is and SM-ie selected by the user is set as the search condition SC _gra. A search using the search condition SC _gra is executed. Therefore, the user can perform a sound material search operation by combining this gradation search with other feature value specification search and similar search, thereby selecting a desired sound material from a large number of sound materials recorded in the sound material databases 25 and 26. The sound material can be searched efficiently.

Although one embodiment of the present invention has been described above, the present invention may have other embodiments. For example, it is as follows.
(1) In the above embodiment, in the search control process, the CPU 21 searches for the feature vector P obtained by extrapolating the feature vector P of the sound material indicated by each of the two selected symbols SM-is and SM-ie. A sound material may be searched as a condition, and a symbol SM-i indicating the searched sound material may be displayed side by side outside the two symbols SM-is and SM-ie.

(2) In the above embodiment, three or more of the symbols SM-i displayed in the search operation area 31 may be selected as search conditions for gradation search. In this embodiment, as shown in FIG. 9, when three symbols SM-i (for example, symbols SM-21, SM-22, and SM-23) displayed in the search operation area 31 are selected. In addition, in the feature vector space, the difference vector between the feature vector P of the sound material indicated by the symbol SM-21 and the feature vector P of the sound material indicated by the symbol SM-22 is one side, and the sound indicated by the symbol SM-21 is displayed. While obtaining a feature vector P distributed in a matrix within a parallelogram having a difference vector between the feature vector P of the material and the feature vector P of the sound material indicated by the symbol SM-23 as one other side, A sound material having the feature vector P closest to the feature vector P is searched, and symbols SM-i indicating each sound material as a search result are displayed in a matrix. Good.

(3) The above aspect (2) may be modified as follows. That is, as shown in FIG. 10, when three symbols SM-i (for example, symbols SM-31, SM-32, and SM-33) displayed in the search operation area 31 are selected, the feature amount is selected. In the vector space, a feature vector P distributed in a triangle having each feature vector P of each sound material indicated by the symbols SM-31, SM-32, SM-33 as vertices is obtained, and the feature vectors A sound material having the feature vector P closest to P may be searched, and the symbol SM-i indicating each sound material as a search result may be displayed in a three-dimensional manner.

(4) In the above embodiment, the Euclidean distance between the feature amount vectors of the sound material is used as an index indicating the similarity between the sound materials . However , a feature amount designation search is performed using another similarity index. Similarity search and gradation search may be performed.

(5) In the above-described embodiment, in gradation search, two symbols SM-i indicating the reference sound material are displayed, and Num sound materials having a feature value similar to the feature value vector are indicated by Num point interpolation. Although the symbol SM-i column is displayed between the two symbols SM-i, the display of the two symbols SM-i indicating the reference sound material may be omitted.

(6) In the above embodiment, the search operation area 31 on the search screen 31 is set to the layer layer-k (k) of k (k = 1 to 16) layers corresponding to each of the frames 5-k (k = 1 to 16). = 1 to 16), and the symbol SM- that fills the frame 5-k from among the symbols SM-i displayed in each layer layer-k in the layer layer-k (k = 1 to 16) of the search operation area 31 i may be selected.

(7) In the above embodiment, the user performs the performance work that continues after filling all the 16 frames 5-k (k = 1 to 16) in the search operation area 31 with the symbols SM-i in the material determination work. Moved on. However, the performance shifts to a performance work in a state where a part of the 16 frames 5-k (k = 1 to 16) is filled with the symbols SM-i, and the symbols SM-i on the part of the frames 5-k indicate. You may enable it to perform using sound material.

DESCRIPTION OF SYMBOLS 10 ... Sound material search apparatus, 11 ... Interface, 90 ... Sound system, 13 ... Operation part, 14 ... Mouse, 15 ... Keyboard, 16 ... Display part, 20 ... Control part, 21 ... CPU, 22 ... RAM, 23 ... ROM 24, hard disk, 25, 26 ... sound material database, 27 ... rhythm pattern database, 28 ... sound material search program.

Claims (2)

Storage means for storing each sound material and a feature vector indicating the characteristics of the sound material in association with each other;
Operation means;
Display means for displaying a symbol indicating sound material;
One or a plurality of sets obtained by interpolating between a plurality of feature amount vectors stored in the storage unit in correspondence with the plurality of reference sound materials, using a plurality of sound materials specified by the operation of the operation unit as a reference sound material Control for searching one or a plurality of sound materials having a feature vector most similar to each of the feature vectors in the storage unit, and displaying a plurality of symbols indicating each sound material as a search result on the display unit Means ,
The control means is similar to the feature vector stored in the storage means in association with the sound material indicated by the selected symbol when a symbol indicating one sound material is selected by the operation of the operation means. A sound material having a feature vector is searched from the storage means for a predetermined upper number, and a symbol indicating the upper predetermined number of sound materials as a search result is displayed in the vicinity of the selected symbol. Sound material search device. Storage means for storing each sound material and a feature vector indicating the characteristics of the sound material in association with each other;
  Operation means;
  Display means for displaying a symbol indicating sound material;
  One or a plurality of sets obtained by interpolating between a plurality of feature amount vectors stored in the storage unit in correspondence with the plurality of reference sound materials, using a plurality of sound materials specified by the operation of the operation unit as a reference sound material Control for searching one or a plurality of sound materials having a feature vector most similar to each of the feature vectors in the storage unit, and displaying a plurality of symbols indicating each sound material as a search result on the display unit Means,
  The control means displays two symbols indicating two reference sound materials on the display means in response to an operation of the operation means, and indicates the two reference sound materials in the display screen of the display means. A sound material searching apparatus, wherein the number of sound materials to be searched from the storage means is determined according to a distance between two symbols.

Images