JP2011118221A - Musical piece creation device and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To automatically create musical piece data of high quality to which expression is imparted, the musical piece data being rich in variety. <P>SOLUTION: In this musical piece creation system, a plurality of reference parts Rp are obtained by dividing a reference section data Rt for reference (RD) to provide them in a phrase data RB and a target section St which is to be arranged in a target song So is divided into a plurality of target parts Tp by referring to performance information included in the section St (PD). A reference part Rp which satisfies a predetermined compatibility condition is retrieved from the database RB for each target part Tp. When the compatible reference part Rp is retrieved, the target part Tp is written again with performance information based on the reference part Rp (PR). When the reference part Rp which is compatible until the middle of the target part Tp is retrieved, it is replaced with the performance information based on the reference part Rp to the middle, and the database RP is retrieved again, for the rest part. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a music generation system capable of automatically generating music data with a variety of expressions and rich variations.

  Conventionally, various music generation apparatuses that automatically generate music data have been proposed. For example, in the automatic composition apparatus disclosed in Patent Document 1, an overall composition condition template is selected, and a melody template suitable for the entire composition condition is searched. The melody template stores data such as the melody skeleton, chord progression, melody pitch, and melody hit points (the hit point is the melody pronunciation timing, and the number of dots in one measure corresponds to the number of notes in one measure). These data are transformed according to user settings. For the dot data (rhythm pattern) in the melody template, the rhythm pattern feature connection data for each measure is read for each measure based on the user setting, the rhythm pattern database is searched, and the rhythm pattern for one measure is found. Selected. Then, the rhythm pattern data of the entire song is generated by connecting the rhythm patterns of each measure throughout the song. In the electronic musical instrument of Patent Document 2, by specifying a chord progression and a key for a melody pattern, a scale corresponding to the chord is selected, and transposition corresponding to the chord is performed for each sound of the melody pattern, The pitch is corrected to match the scale.

  In the music generation device of Patent Literature 3, a plurality of phrases can be extracted from existing music and converted into a database, and the music can be easily created by rearranging the phrases using this. Here, when a phrase is extracted from an existing musical piece, a unique attribute is assigned to each phrase from which a key, chord, first sound, last sound, and the like are extracted, and a database is created. Furthermore, in the melody generation method of Patent Document 4, a phrase database for music generation is created, and phrase data is cut out from musical score information. There are two types of phrase data extraction: fixed-length extraction and variable-length extraction. In the case of fixed-length, the phrase data is extracted by shifting the extraction position by one measure at a length of 1 to 4 measures. In this case, the musical score information is cut out between rests, and when the extracted phrase data is less than one measure, it is combined with the shorter one of the phrase data before or after the extracted phrase data, or cut out When the phrase data is 4 bars or more, the phrase data is divided after the longest note value in the bar.

JP 2000-99015 A Japanese Patent Laid-Open No. 6-2222764 JP 2006-178104 A JP 2007-219139 A

  However, the automatic composition apparatus that designates a lot of composition conditions as in Patent Document 1 is not easy and basically requires complicated setting and editing operations in units of measures. The electronic musical instrument of Patent Document 2 has a goal of determining the hit point, pitch, and pitch, and cannot expect generation of high-quality performance data such as facial expression. Moreover, in the music production | generation apparatus of patent document 3, when the phrase extracted from the existing music is connected, there is a possibility that originality may be lacking in the completion. In the phrase extraction method disclosed in Patent Document 4, all of them are one measure or more, and there is a possibility that the originality is lacking in the finished product. Furthermore, even if these conventional techniques are combined, it is difficult to generate high-quality music data with originality by a simple operation.

  In view of such circumstances, the present invention adds a facial expression by arranging and sequentially synthesizing optimum phrase data for each phrase section disassembled with reference to the phrase database for the arrangement designation section in the original music data. It is an object of the present invention to provide a music generation system capable of automatically generating high-quality music data rich in variations.

  According to the main feature of the present invention, a plurality of phrase data (reference parts Rp:..., Rpb,..., Rpa) divided from performance data (reference section data Rt: Rtj; Rtk) referenced for music generation. ,..., A target section specifying means (TS) for specifying a music generation target section (target section St) for the original music data (target song So), and a target section specifying means (TS Phrase division that divides the music generation target section (St) into a plurality of phrase sections (target parts Tp: Tpa, Tpb,...) With reference to performance information included in the music generation target section (St) specified by The phrase section (Tp) divided by the means (PD; P1 to P5) and the phrase dividing means (PD; P1 to P5) As a result of the search by the phrase search means (PS; P6) and the phrase search means (PS; P6) for searching the phrase data (Rp) that matches the hit point from the phrase storage means (RB), the phrase section (Tp) First performance information conversion means for writing performance information generated based on the retrieved phrase data (Rp) into the phrase section (Tp) when the phrase data (Rp) conforming to is retrieved (P7 = YES) (PR; P8) and when the phrase data (Rp) suitable for the middle of the phrase section (Tp) is searched as a result of the search by the phrase search means (PS; P6) (P7 = NO, P9 = (YES), and the performance information generated based on the searched phrase data (Rp) is written in the middle part of the phrase section (Tp) ( 10) For the remaining part of the phrase section (Tp), the phrase data (Rp) matching the hit point is searched from the phrase storage means (RB) (P13), and corresponding performance information is generated according to the search result. And a musical piece generating apparatus (Claim 1) provided with second performance information converting means (PS, PR; P9 to P13) to be written in the remaining portion (P10), and performance data referred to for musical piece generation (Phrase storage means (RB) for storing a plurality of phrase data (reference parts Rp:..., Rpb,..., Rpa,...) Divided from (reference section data Rt: Rtj; Rtk), and music generation A target for specifying a music generation target section (target section St) for the original music data (target song So) to a computer functioning as a device The music generation target section (St) is referred to as a plurality of phrase sections (targets) with reference to performance information included in the section specifying step (TS) and the music generation target section (St) specified in the target section specifying step (TS). The hit points are suitable for each of the phrase division steps (PD; P1 to P5) divided into parts Tp: Tpa, Tpb,..., And the phrase sections (Tp) divided in the phrase division steps (PD; P1 to P5). A phrase search step (PS; P6) for searching the phrase data (Rp) from the phrase storage means (RB), and a phrase data (phrase data (Tp) adapted to the phrase section (Tp) as a result of the search in the phrase search step (PS; P6)) Rp) is retrieved (P7 = YES), the performance information generated based on the retrieved phrase data (Rp) As a result of the search in the first performance information conversion step (PR; P8) and the phrase search step (PS; P6) to be written in the phrase interval (Tp), the phrase data (Rp) adapted to the middle of the phrase interval (Tp) ) Is retrieved (P7 = NO, P9 = YES), the performance information generated based on the retrieved phrase data (Rp) is written to the middle part of the phrase section (Tp) (P10). For the remaining portion of the phrase section (Tp), the phrase data (Rp) matching the hit point is searched from the phrase storage means (RB) (P13), and corresponding performance information is generated according to the search result. A musical tone generation program for executing a procedure comprising a second performance information conversion step (PS, PR; P9 to P13) to be written in the remaining part (P10). [Claim 3] is provided. Note that the parentheses indicate reference symbols, terms, and the like of the embodiments added for convenience of understanding, and the same applies to the following.

  In the music generating device according to the present invention, in each of the plurality of phrase sections (Tp) divided by the phrase dividing means (PD; P1 to P5), note state information (chord state) indicating a note state for each hit point. Is recorded (P4), and each of the plurality of phrase data (Rp) stored in the phrase storage means (RB) is recorded with note state information (chord state) indicating the state of the note at each hit point (R4). The phrase searching means (PS; P6) and the second performance information generating means (PS, PR; P9 to P13) check not only the hit points but also the note state information (chord state). In the music generation program according to the present invention, each of the plurality of phrase sections (Tp) divided in the phrase dividing step (PD; P1 to P5) is used. Is recorded with note state information (chord state) indicating the state of the note for each hit point (P4), and each of the plurality of phrase data (Rp) stored in the phrase storage means (RB) includes a note for each hit point. Note state information (chord state) is recorded (R4), and the phrase search step (PS; P6) and second performance information generation step (PS, PR; P9 to P13) It can be configured to check the conformity of the note state information (chord state).

  In the music generation system according to the main feature of the present invention (Claims 1 and 3), a plurality of quality data (reference section data Rt: Rtj; Rtk) referred to for music generation are divided (decomposed). Phrase data (reference parts Rp:..., Rpb,..., Rpa,...) Are prepared in the phrase storage means (RB). With respect to the original music data (target song So), when a music generation target section (target section St), that is, a section to be arranged is designated by user operation (TS), performance information included in the designated music generation target section (St) is displayed. The music generation target section (St) is divided (decomposed) into a plurality of phrase sections (target parts Tp: Tpa, Tpb,...) (PD; P1 to P5). Next, for each divided phrase section (Tp), the hit points (rhythm positions) match (match) [Here, “match” includes at least all rhythm positions (= “completely includes”) Phrase search (PS; P6) for searching (searching) the phrase data (Rp) from the phrase storage means (RB) is performed, and as a result of the phrase search (PS; P6), the phrase section When phrase data (Rp) conforming to (Tp) is searched (P7 = YES), performance information generated based on the searched phrase data (Rp) is written in the phrase section (Tp) (overwriting) Then, the performance information of the phrase section (Tp) is rewritten (PR; P8). On the other hand, as a result of the phrase search (PS; P6), when only the phrase data (Rp) that fits in the middle of the phrase section (Tp) is searched (P7 = NO, P9 = YES), the search is performed. The performance information generated based on the phrase data (Rp) is written (overwritten) in the middle part of the phrase section (Tp) (P10), and the remaining points of the phrase section (Tp) match. The phrase data (Rp) is re-searched (re-searched) from the phrase storage means (RB) (P13), and corresponding performance information is generated and written in the remaining part according to the result of the re-search (P10).

  In other words, in the present invention, for example, high-quality performance data (Rt) that has been subjected to facial expression or the like is divided into a plurality of phrase data (Rp) in advance and prepared in the phrase database (RB), and the original music data ( So) the designated section (St) is decomposed into a plurality of phrase sections (Tp), and the phrase database (RB) is referred to for each phrase section (Tp), and the phrase is used as a matching condition whether or not the rhythm positions match. When the data (Rp) is searched and the phrase data (Rp) having the same rhythm position (including at least all rhythm positions) in each phrase section (Tp) is searched (P7 = YES), the searched phrase data When (Rp) is adopted for the phrase section (Tp) and the rhythm position does not completely match the entire phrase section (Tp) (P7 = O, P9 = YES), the phrase portion having the same rhythm position is adopted until the middle of the phrase section (Tp), and the re-search (re-search) is performed for the remaining portion after the middle (P13). Thus, the designated section (St) of the original music data (So) is automatically arranged with reference to the phrase database (RB) (AR).

  Therefore, according to the present invention, the optimal phrase data that matches the rhythm position for each phrase section is divided from the plurality of phrase data prepared in the phrase database for the specified section of the original music data. By selecting and sequentially synthesizing, a new phrase can be generated, and high-quality performance data with a facial expression can be easily and automatically generated. Here, since the newly generated phrase is not simply code-corresponding to the phrase data prepared in advance, it is possible to compose music rich in variations. Also, since there are many different types of phrase data, ranging from several measures to short ones less than one measure, it is possible to generate songs with various combinations and originality. Furthermore, since it is possible to adopt the middle of the phrase regardless of the length of the phrase, the prepared phrase data can be utilized without waste.

  In the music generating system according to the present invention (Claims 2 and 4), note state information (chord state) indicating the state of a note for each hit point is recorded in each of a plurality of divided phrase sections (Tp) ( P4), note state information (chord state) indicating the state of the note for each striking point is also recorded in each of the plurality of phrase data (Rp) stored in the phrase storage means (RB) (R4), and the phrase search ( In the case of P6) or the re-search of the remaining part (P13) (PS), in addition to the matching (matching) of the hit point (rhythm position), the matching (matching) of the note state information (chord state) is checked as the matching condition. Like to do. Therefore, according to this invention, the optimal phrase data corresponding to the musical feature of each phrase section in the specified section of the original music data can be selected from the phrase database.

1 shows a configuration example of a music generation system according to an embodiment of the present invention. It is a figure for demonstrating the principle of the music production | generation by one Example of this invention. It is a figure for demonstrating the example of the chord state by one Example of this invention. It is an example of a reference registration process flow by one Example of this invention. It is a part of example of a phrase conversion process flow by one Example of this invention. It is the other part of the example of a phrase conversion process flow by one Example of this invention. It is a reference adoption processing flow example by one Example of this invention.

[Overview of system configuration]
FIG. 1 shows a configuration example of a music generation system according to an embodiment of the present invention, and FIG. 1 (1) shows a hardware configuration example of a music generation device according to an embodiment of the present invention. A music generation device according to an embodiment of the present invention is a kind of computer having an electronic music information processing function, and for example, an electronic music device such as an electronic musical instrument is used and functions as a music generation device in a music generation mode. Thus, a music generation system is constructed. As shown in FIG. 1 (1), the music generating device includes a central processing unit (CPU) 1, a random access memory (RAM) 2, a read only memory (ROM) 3, an external storage device 4, an input operation unit 5, a display. Unit 6, sound source 7, communication interface (I / F) 8, and the like, and these elements 1 to 8 are connected to each other via a bus 9.

  The CPU 1 constitutes a data processing unit together with the RAM 2 and the ROM 3, and executes various music information processing including music generation processing according to a predetermined control program including a music generation processing program called “phrase conversion processing”. The RAM 2 is used as a work area for temporarily storing various data necessary for these processes. For example, in the music generation mode, the RAM 2 is set from a phrase data storage unit (phrase database) RB provided in the external storage device 4. Accordingly, a phrase data holding area for holding a phrase data group (Rp) selectively read out is provided. In addition, various control programs and various control data necessary for executing these processes are stored in the ROM 3 in advance.

  The external storage device 4 includes a storage medium such as a semiconductor memory such as an HD (hard disk), an FD (flexible disk), a CD (compact disk), a DVD (digital multipurpose disk), and a flash memory, and a driving device thereof. Arbitrary information such as a control program and various automatic performance data can be stored in an arbitrary storage medium, and a phrase database RB for storing and managing various phrase data (Rp) in a predetermined storage medium Is built. Further, the storage medium may be detachable or may be built in the music generation device, and the detachable recording medium includes a USB memory.

  The input operation unit 5 introduces operation information corresponding to the detected contents to the data processing units (1 to 3) by detecting performance operations such as a keyboard and setting operations such as a key switch and operations of these operations. And an operation detection circuit. The data processing unit controls each unit of the music generation device based on the operation information, for example, sends performance information corresponding to the performance operation input to the sound source 7 according to the operation of the performance operator, or mode setting in the setting operator It can be operated in the music generation mode by operating the buttons. The display unit 6 includes a display such as an LCD for displaying various information necessary for performance and setting, a built-in lamp attached to the switches in the input operation unit 5, and the like. Controls with commands, provides display support for performance and setting operations.

  The sound source 7 includes a sound source unit and a DSP, generates a musical sound signal based on actual performance information obtained from the performance operation information from the input operation unit 5, or generates automatic performance data from the storage device 4 or the communication I / F 8. A musical tone signal based on the generated musical tone signal is generated, and a predetermined effect is given to the generated musical tone signal. The sound system 10 connected to the sound source 7 includes a D / A converter, an amplifier, a speaker, and the like, and generates a musical sound based on a musical sound signal output from the sound source 7.

  The communication I / F 8 includes a music wired I / F such as MIDI, a general-purpose network I / F such as USB, or a general-purpose short-range wireless I / F such as wireless Lan, and can communicate with the external electronic music apparatus MM. For example, music information such as MIDI performance data and phrase data (Rp) is stored in the external storage device of the external electronic music device MM. Music information can be used in this music generation device.

[Outline of music generation system]
FIG. 1 (2) is a functional block diagram of a music generation system according to an embodiment of the present invention. In the music generation mode, the data processing units (1 to 3) including the CPU 1 of the music generation device (electronic music device), as shown in FIG. 1 (2), the target section specifying unit TS, the arrangement unit AR, and the reference phrase. The phrase data storage unit RB that functions as the division unit RD and is provided in the external storage device 4 of the music generation device (electronic music device) stores a wide variety of phrase data Rp called “reference parts”. A database is built. The reference phrase database is also simply referred to as a phrase database, and is described with the same reference symbol “RB” as the phrase data storage unit.

  The phrase data Rp is called “target song” or “target song data” and is created from high-quality reference song data Rs that is referred to when the new song data Soa is generated from the original song data So that is the source of the song generation. can do. The reference song data Rs is also referred to as reference song or reference song data. When creating phrase data, first, reference zone data is extracted from the reference song data Rs by the reference segment extraction unit (not shown). Rt (referred to as reference section data) is cut out as reference performance data. Then, the reference phrase dividing unit RD divides (decomposes) the reference interval data Rt into a plurality of reference phrase intervals Rp according to the contents of performance information such as notes and rests included in the reference interval data (reference performance data) Rt. Data of each divided reference phrase section Rp is stored as phrase data (reference parts) in the phrase database RB (the reference phrase “RB” is added to the reference phrase section), and the phrase database RB is A large number of phrase data Rp are organized and stored, for example, by reference music data Rs.

  With respect to the target song (original music data) So, first, the target section designating unit TS designates and designates a music generation target section, that is, a section (target section) to be arranged according to a target section specifying operation by the user. Data of the music generation target section (called target section data) St is introduced into the arrangement unit AR.

  The arrangement unit AR is mainly composed of a target phrase division unit PD, a phrase search unit PS, and a data conversion unit PR, and each functional unit PD, PS, PR has a preset resolution (quantized length or quantized value). The corresponding function is executed according to parameters Pr such as rest length (rest section length), maximum measure length, key, time signature, tempo, chord, and the like. The target phrase dividing unit PD divides the target section data St into a plurality of target phrase sections (also simply referred to as phrase sections) Tp according to the contents of performance information such as notes and rests included in the input target section data St. (Decomposition), and for each divided target phrase section Tp, a rhythm position (batting point) and a “chord state” indicating the state of a note (pitch designation information) at each rhythm position are obtained, and the obtained rhythm The data of each target phrase section Tp to which the position and chord state information is added is output as a “target part” to the phrase search unit PS. In addition, the same reference symbol “Tp” as the target phrase section is appended to the target part.

  The phrase search unit PS is predetermined for each divided target phrase section, that is, for each target part Tp, with respect to a predetermined phrase data group (reference parts group) in the phrase database RB specified in accordance with a search target specifying operation by the user. Phrase data that meets the above condition, that is, the reference part Rp is searched (searched), and the data of the searched reference part Rp is sent to the data converter PR together with the data of the target part Tp. In this search process, information on the rhythm position (batting point) and chord state is used as the matching condition (search condition), and the rhythm positions of the respective target parts Tp match, that is, at least all the rhythm positions of the target parts Tp. Is included in the rhythm position of the reference part Rp (referred to as “completely including the rhythm position”), and the reference part Rp in which the chord state of the matching rhythm position (ie, included rhythm position) matches is a phrase A search is sequentially performed from the database RB.

  The data converter PR is also referred to as a note adjuster, so that each pitch of the searched reference part Rp matches the chord and key of the target part Tp, and matches the original range of the target part Tp. Performing note adjustment such as changing, thereby generating performance information suitable for the target part Tp based on the retrieved reference part Rp, and sequentially rewriting the performance information of each target part Tp with the generated performance information. Go. At the same time, expression data (control data) included in the retrieved reference part Rp is written in the target part Tp. Then, by rewriting the target part Tp in the entire target section data St, the arranged target section data Sta is generated, and further, the arranged target section data Sta is incorporated into the target section St of the original music data So. New song data with ad lib (new music data) Soa is generated.

  FIG. 2 is a diagram for explaining in principle the mechanism of music generation according to one embodiment of the present invention. The target section data St in the target song So is a target phrase as shown in FIG. The dividing unit PD divides the target parts Tp:..., Tpa, Tpb,. On the other hand, the phrase search unit PS selects a desired reference song Rs according to the reference data selection operation by the user, and the selected reference song Rs is selected from the phrase database RB as shown in FIG. The reference part (group) Rp included in the two reference section data Rtj and Rtk stored corresponding to is loaded into the phrase data holding area of the RAM 2. When the phrase search unit PS is instructed to start a phrase search in response to a search start operation by the user, the phrase search unit PSs sequentially specifies the target parts Tp in the target section data St from the top (not shown). The reference part Rp that matches the specified target part Tp is sequentially searched from the reference section data Rtj and Rtk.

  For example, for a certain target part Tpa, the reference section data Rtj and Rtk are searched according to the matching conditions. As a result, as shown in the lower part of FIG. 2 (2), the reference corresponding to the target part Tpa from the reference section data Rtk. When the part Rpa is found, the target part Tpa is replaced with the corresponding reference part Rpa. The same search is performed for the next target part Tpb. As a result, when the reference part Rpb is found in the reference interval data Rtj as shown in the upper part of FIG. 2 (2), the target part Tpb is replaced with the corresponding reference part Rpb. To do. In this way, by sequentially synthesizing phrases with reference to reference parts (phrase data) Rp:..., Rpa, Rpb,... Corresponding to each target part Tp:..., Tpa, Tpb,. New target section data Sta can be generated.

  In the target phrase division, the target section data St is divided into a plurality of target parts (phrase sections) Tp based on rest information or the like. That is, as indicated by “Δ” in FIG. 2 (1), the target part (for example, Tpa) is located at a position where there is a rest section of a predetermined length in the target section data St or a position (Δ) at which a predetermined maximum bar length is reached. , Tpb). After the division, each target part Tp is quantized and analyzed with a predetermined resolution (for example, quarter note length) designated by the user setting to obtain rhythm position and chord state information.

  The chord state is information representing the state of the note (pitch designation information) at each striking point (rhythm position), and FIG. 3 shows an example of the chord state according to one embodiment of the present invention. In this example, there are five types of chord states with state values “1” to “5”. A chord state with a state value = “1” is called a “single note” and has one note-on at the rhythm position. A chord state with a state value = “2” is called “quick playing”, and there are two or more note-ons at the rhythm position, at least one of which is 1 / of the quantization length (resolution). This represents a state in which the sound length is less than 2. A chord state with a state value = “3” is called “chord performance”, and there are two or more note-ons at the rhythm position, all of which have a tone length that is ½ or more of the quantize length. Represent. Further, a chord state with a state value = “4” is called “pitch change by PB (single)”, and there is only one pitch change by pitch bend (PB) at the rhythm position. A chord state having a state value = “5”, which represents a non-existing state, is called “pitch change by PB (Multi)”, and there is at least one pitch change by pitch bend (PB) at the rhythm position. This represents a case where there is one or more pitch changes due to normal note-on or pitch bend (PB). Note that the state value may be set more finely, for example, in the case where the state value = “3”, two note-on cases, and three note-on cases.

  In the phrase search (PS), the reference part Rp is searched on the condition that the rhythm position and the chord state coincide with the target part Tp. At that time, the search for the corresponding reference part Rp by matching the chord states is performed on condition that the state values of the chord states are the same value.

  In the phrase search (PS) of each target part Tp, the reference part Rp that perfectly matches the target part Tp is not searched, and if the reference part Rp that matches partway through the target part Tp is searched, data conversion is performed. The part PR generates performance information suitable for the middle of the target part Tp based on the retrieved reference part Rp, and rewrites the middle of the target part Tp with the performance information. For the remaining part of the target part Tp after rewriting, the reference part Rp that matches the rhythm position and chord state is re-searched (re-searched) from the reference part group held in the phrase holding area, and re-searched. According to the search result, corresponding performance information is generated and the remaining portion is rewritten. Even in the re-search, when the matching reference part Rp is searched only halfway through the remaining part of the target part Tp, the performance information corresponding to the matching reference part Rp is used until the middle of the remaining part of the target part Tp. The same processing is repeated for the rewriting and thereafter.

  In addition, in the phrase search (PS) of each target part Tp, if neither the reference part Rp that matches the target part Tp nor the reference part Rp that matches the target part Tp is searched, the leading sound of the target part Tp The performance information at the (note-on) position is set to “unsupported”, and the second sound (note-on) position and subsequent positions are re-searched (re-searched). Further, if no suitable reference part Rp is found even after re-searching, the performance information of the re-searched first note (note-on) is set to “unsupported”, and the same processing is performed for the next sound (note-on) and thereafter. repeat. Here, with respect to the performance information in the target part Tp that is set to “unsupported”, some performance information is generated separately, for example, the musical information is left as it is.

  As described above, in the music generation system according to one embodiment of the present invention, the high-quality reference section data (performance data) Rt referred to for music generation is divided into a plurality of reference parts (phrase data) Rp ( RD) prepared in a phrase database (phrase data storage unit) RB. A target section (song generation target section) St to be arranged for the target song (original music data) So is divided into a plurality of target parts (phrase sections) Tp by referring to the performance information included in the section St (PD) For each target part Tp, a phrase search PS for searching (searching) a reference part Rp satisfying a predetermined matching condition from the phrase database RB is performed. As the matching conditions in the phrase search PS, it is necessary that at least the hit points (rhythm positions) match, and it is preferable that the note states at the respective hit points called “chord states” match. As a result of the phrase search PS, when a reference part Rp suitable for the target part Tp is searched, the target part Tp is rewritten with performance information generated based on the searched reference part Rp (PR). On the other hand, as a result of the phrase search PS, when only the reference part Rp that matches the target part Tp is searched, the part of the target part Tp is generated based on the searched reference part Rp. The reference part Rp that satisfies the predetermined matching condition is re-searched (re-searched) from the phrase database RB for the remaining part after the performance information. Therefore, according to this system, the portion St desired to be arranged in the target song So is converted by the reference part Rp having a good quality that approximates the rhythm and syllable for each target part Tp, and a desired ad-lib portion can be created. .

[Example of processing flow]
FIG. 4 is a flowchart showing an example of the operation of reference registration processing according to one embodiment of the present invention. In this music generation device, when operating in an arbitrary operation mode such as a standby mode, an actual performance mode, or an automatic performance mode, the operation mode of the reference registration mode can be set by operating the mode setting button. In the reference registration mode, the reference section data (reference performance data) Rt in the reference song (reference song data) Rs is stored in the RAM 2 in a predetermined manner prior to the music generation process (phrase conversion process) by the reference registration process shown in FIG. The analysis result can be loaded into the storage area and the analysis result can be stored in the phrase database RB as a reference part Rp.

  In FIG. 4, first, in step R <b> 1, the CPU 1 can evaluate (analyze) the performance information included in the reference section data Rt as a base in order from the first note (having at least timing and pitch information). Read and specify (determine) resolution by user operation. In the subsequent step R2, every time a rest section [“Δ” in FIG. 2 (2)] longer than the length (designated rest length) designated by the user operation is detected in the reference section data Rt, at that position. Divide into reference parts Rp. The rest length is preferably analyzed with a plurality of predetermined lengths. In addition, for the section width of the reference part Rp to be divided, an upper limit is specified in advance as a maximum measure length such as two bars, and a rest section is detected and divided so as not to exceed the upper limit (maximum measure length). It is good to go. That is, based on the rest information included in the reference section data Rt, when a rest section of a specified length or more is detected, the reference section Rp is divided and the length of the divided reference part Rp is designated. The maximum bar length is limited to the maximum bar length (for example, two bar lengths) at the longest.

  After the reference part division in step R2, the processing of steps R3 and R4 is sequentially performed on the performance information included in the plurality of divided reference parts Rp in order from the top reference part Rp. That is, in step R3, the specified resolution for each reference part Rp (referred to as the specified resolution. For example, 4 minutes, 4 minutes, 3 stations, 8 minutes, 8 minutes, 3 stations, 16 minutes, 16 minutes, 3 stations, etc.) Quantize each to obtain a rhythm position (a plurality of types of quantization may be performed in advance), and obtain rhythm position information representing the rhythm position of each calculated note in the predetermined storage area in the RAM 2 and the original reference part Rp. Hold in association. In step R4, a chord state at each obtained rhythm position is obtained, and chord state information representing the obtained chord state as a state value is stored in a predetermined storage area in the RAM 2 in association with the rhythm position information. In the next step R5, it is determined whether or not the processing for obtaining the rhythm position information and chord state information has been completed for all the reference parts Rp in the reference section data Rt, and the processing of all the reference parts Rp has not been completed. (R5 = NO), the process returns to step R3, and the processing of steps R3 and R4 is repeated for the next reference part Rp.

  When the processing is completed for all the reference parts Rp (R5 = YES), the process proceeds to step R6, and the rhythm position information and chord state information obtained in steps R3 and R4 and stored in the RAM 2 for each reference part Rp. Is stored in the phrase database RB and registered as reference parts Rp for each reference song Rs and for each reference section data Rt. Then, the reference registration process is terminated, and the original operation mode is restored. In addition, when registering in the phrase database RB, it is preferable that each reference part Rp is registered for a plurality of types according to the quantize length and rest length used in the analysis.

  5 to 7 are flowcharts showing an operation example of phrase conversion processing (music generation processing) according to one embodiment of the present invention. When the mode setting button is operated to enter the phrase conversion mode operation state, the target section data St in the target song (original music data) So is stored in a predetermined storage area in the RAM 2 by the phrase conversion processing shown in FIGS. The new target section data Sta can be generated while referring to the reference part Rp of the phrase database RB.

  5 to 6, the CPU 1 first analyzes the target section data St by the same procedure as the steps R1 to R5 of the reference registration process (FIG. 4) described above by the analysis process of steps P1 to P5. The result can be used as the target part Tp for the search and conversion processing in steps P6 to P15. That is, in step P1, the performance information of the target section data St is read so that it can be evaluated (analyzed) sequentially from the first note (having at least timing and pitch information), and the resolution is designated (determined) by the user operation. . In the next step P2, a rest section [“Δ” in FIG. 2 (1)] of a specified length or more is detected based on rest information included in the target section data St, and divided into target parts Tp of the maximum measure length or less. I will do it. That is, as in step R2, when a rest interval of a specified length or longer is detected, or even if such a rest interval is not detected, a maximum bar length specified in advance (for example, two bar lengths). Is divided into target parts Tp. Next, in step P3, each target part Tp is quantized with a designated resolution to obtain rhythm position information (quantization may be performed in advance), and in subsequent step P4, each rhythm in the target part Tp. A chord state (state value) at the position is obtained and stored in a predetermined area of the RAM 2 in association with the rhythm position information. And while it determines with the analysis process not being complete | finished about all the reference parts Rp by step P5 (P5 = NO), the process of step P3, P4 is repeated. When analyzing the target section data St, it is quantized with any one of the specified resolutions specified by the user operation among a plurality of resolutions (P3), and the length of the rest section is also determined in advance. The phrase is divided by any one type of rest length designated by the user operation from the plurality of rest lengths (P4). Further, every time the rest section exceeds the maximum measure length, it may be divided as a rest-only section.

  After the target section data analysis process of steps P1 to P5 (P5 = YES), the search process of step P6 is performed, and the rhythm position and the chord state match for each target part Tp specified by the phrase section specifying means PSs. A reference part, that is, a reference part Rp that completely includes the rhythm position and has the same chord state corresponding to the included rhythm position is searched for. In this case, a reference part group corresponding to the designated target song So and the same resolution and rest length as the resolution and rest length used in the target section data analysis processing is set as a search target range, and the reference part group Is read from the phrase database RB and loaded into the phrase data holding area of the RAM 2. Therefore, for example, when the designated rest length serving as a reference is changed in step P2, the reference part group corresponding to the rest length after the change is selected as the search target range. In addition, when the target part Tp is only a rest, any reference part may be adopted according to the setting.

  In the next step P7, as a result of the search process, the reference parts Rp that completely match, that is, “each target part Tp includes at least completely the rhythm position information over all phrase sections (all rhythms of the target part). When a reference part Rp having the same (matching) chord state corresponding to each included rhythm position is found and a completely matching reference part Rp is found. (P7 = YES), a reference adoption process is performed in step P8, and the found reference part Rp is adopted as conversion data for the target part Tp.

  On the other hand, when the reference part Rp that completely matches (= over all phrase sections) is not found (P7 = NO), the process proceeds to step P9 (FIG. 6), and the rhythm position including the chord state matches to the middle of the phrase section. It is determined whether or not a reference part Rp, that is, a reference part Rp in which the rhythm positions coincide with each other and the chord states corresponding to the coincident rhythm positions coincide with each other is found (S9 = YES). Then, the process proceeds to Step P10, where the reference adoption process is performed halfway, and the found reference part Rp is employed as conversion data in the matching section (from the beginning to the middle) of the target part Tp. When a plurality of reference parts Rp are found, the one that matches in a long section (up to a later position) is adopted. On the other hand, if even a reference part that matches partway is not found (P9 = NO), the process proceeds to step P11, where the target part Tp or the “remaining section” that is currently subject to matching is determined. For the target part part, the first note is determined as “unsupported”, the second and subsequent sounds from the beginning are newly determined as “remaining sections”, and if there are no second and subsequent sounds, “no remaining sections” are selected. decide.

  After the process of step P10 or step P11, it is determined in step P12 whether or not there is a “remaining section”. If there is a “remaining section” (S12 = YES), the process proceeds to step P13. ”Is searched for a corresponding reference part. After the remaining section search process in step P13, the process returns to step P9, where the rhythm positions including the chord state are all or halfway matched, that is, the rhythm positions are all or halfway matched, and the rhythm positions are the same. It is determined whether or not a reference part Rp corresponding to each chord state corresponding to is found, and the process of step P10 or step P11 is performed according to the determination result. While there is a remaining section (S12 = YES), step P13 is performed. , P9 to P12 are repeated. For the sound determined to be “unsupported” in step P11, for example, as described above, the original note information may be left as it is, or it may be continued for a period of “unsupported”. If so, the processing may be performed in the same manner as the “rest-only section”.

  After the reference adoption process in step P8 (FIG. 5) or after determining that there is no “remaining section” in step P12 (FIG. 6) (S12 = NO), in step P14 (FIG. 5), the target part Tp It is determined that the search and conversion process has been completed, and the process proceeds to the next step P15. In step P15, it is determined whether or not the search and conversion process has been completed for all target parts Tp. If the search and conversion process has not been completed (S15 = NO), the process returns to step P6, and phrase section specifying means The processing of steps P6 to P15 is repeated for the next target part Tp specified by PSs. When the search and conversion process is completed for all target parts Tp (S15 = YES), the phrase conversion process (music generation process) is terminated and the original operation mode is restored.

  FIG. 7 is a flowchart showing an example of a reference adoption process according to an embodiment of the present invention. The reference adoption process is performed in step P8 (FIG. 5) or step P10 (FIG. 6) of the phrase conversion process (music generation process). Executed. In the first step S1 of the reference adoption process, the rhythm position is completely included in the found reference part Rp for the target part Tp currently designated as the processing target by the phrase section specifying means PSs. Create an event list (performance information) for the sections where the corresponding chord states match. In the next step S2, an event list (performance information) is created for the entire reference part Rp or the matching section up to the middle. At this time, even if outside the section, if there is valid information such as valid control data in the note information within the section, it is added together in the event list.

  In subsequent step S3, note information in the event list created from the reference part Rp is adjusted based on the event list created from the target part Tp. For example, the note pitch of the reference part Rp is changed so as to match the chord information and key information of the target part Tp and so as to match the original range of the target part Tp. In the next step S4, the performance information is deleted for the matching section of the target part Tp, and the performance information is copied from the event list of the reference part Rp in which the note information is adjusted. Furthermore, control data other than the note event data (for example, main volume, aftertouch, etc.) is also copied. After the copy process, the reference adoption process is terminated, and the process returns to step P14 (FIG. 5) or step P12 (FIG. 6) of the phrase conversion process (music generation process).

[Various Embodiments]
The preferred embodiment of the present invention has been described above with reference to the drawings. However, this is merely an example, and the present invention can be variously modified without departing from the spirit of the invention. For example, in the embodiment, the reference part (Rp) is registered in the phrase database in advance, and the target song (St) in the specified section is analyzed every time the music is generated and is used as the target part (Tp). The reference song (Rp) in the section may be analyzed each time to create a reference part. In addition, the reference music data (Rs) that is referred to when the music is generated is one song in the embodiment, but may be a plurality of songs.

Rs Reference song [data] (reference song data),
Rt: Rtj, Rtk Reference section data (reference performance data),
RB [Reference] Phrase database or phrase data storage unit,
Rp: Rpa, Rpb Reference parts (phrase data),
So target song [data] (original song data),
St target section [data] (music generation target section or arrangement section [data]),
Tp: Tpa, Tpb Target parts (phrase interval [data]).

Claims (4)

Phrase storage means for storing a plurality of phrase data divided from performance data referred to for music generation;
A target section specifying means for specifying a music generation target section for the original music data;
Phrase dividing means for dividing the music generation target section into a plurality of phrase sections with reference to performance information included in the music generation target section specified by the target section specifying means;
For each phrase section divided by the phrase dividing means, a phrase searching means for searching phrase data matching phrase data from the phrase storage means;
As a result of the search by the phrase search means, when phrase data suitable for the phrase section is searched, first performance information conversion means for writing performance information generated based on the searched phrase data in the phrase section;
As a result of the search by the phrase search means, when phrase data suitable for the middle of the phrase section is searched, the performance information generated based on the searched phrase data is written to the middle part of the phrase section. The phrase storage means searches for the phrase data that matches the hit point for the remaining part of the phrase section, and according to the search result, the corresponding performance information is generated and the second performance information conversion means for writing in the remaining part. A music generation device comprising the music generation device. In each of the plurality of phrase sections divided by the phrase dividing means, note state information indicating a note state is recorded for each hit point,
In each of the plurality of phrase data stored in the phrase storage means, note state information indicating the state of the note for each hit point is recorded,
2. The music generating apparatus according to claim 1, wherein the phrase searching unit and the second performance information generating unit check the matching of the note state information in addition to the matching of the hit points. A computer comprising phrase storage means for storing a plurality of phrase data divided from performance data referred to for music generation, and functioning as a music generation device,
A target section specifying step for specifying a music generation target section for the original music data;
A phrase dividing step of dividing the music generation target section into a plurality of phrase sections with reference to performance information included in the music generation target section specified in the target section specifying step;
For each phrase section divided in the phrase dividing step, a phrase searching step for searching phrase data with matching hit points from the phrase storage means;
As a result of the search in the phrase search step, when phrase data suitable for the phrase section is searched, a first performance information conversion step for writing performance information generated based on the searched phrase data in the phrase section;
As a result of the search in the phrase search step, when the phrase data suitable for the middle of the phrase section is searched, the performance information generated based on the searched phrase data is written to the middle of the phrase section. From the second performance information conversion step, the phrase data corresponding to the hit point is searched from the phrase storage means for the remaining portion of the phrase section, and corresponding performance information is generated and written in the remaining portion according to the search result. A musical sound generation program that executes a procedure consisting of In each of the plurality of phrase sections divided in the phrase dividing step, note state information indicating the state of the note for each hit point is recorded,
In each of the plurality of phrase data stored in the phrase storage means, note state information indicating the state of the note for each hit point is recorded,
The musical tone generation program according to claim 3, wherein in the phrase search step and the second performance information generation step, in addition to the hit point match, the note state information match is checked.

Images