JP2013101288A - Musical piece generation device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To generate a musical piece in which a character string such as lyrics properly corresponds to each unit section of the musical piece.SOLUTION: A musical piece generation device 100A generates a musical piece in which each of a plurality of unit sections is applied with a melody. A storage device 24 stores a plurality of pieces of melody data DM designating different melodies. An information acquisition part 32 acquires a designated character string DL composed of a plurality of characters. A musical piece generation part 34 distributes a plurality of character groups obtained by dividing the designated character string DL to each unit section, and calculates a melody evaluation value in accordance with the number of characters of each character group corresponding to each of the plurality of unit sections and the number of notes of the melody designated by the melody data DM for each melody data DM. The musical piece generation part 34 selects any of the plurality of melody data DM in accordance with the melody evaluation value of each melody data DM in each of the plurality of unit sections.

Description

  The present invention relates to a technique for generating music.

  A technique for automatically generating music (automatic music technique) has been proposed. For example, Non-Patent Document 1 discloses a musical composition defined by a note sequence selected so that the pitch varies according to the prosody of lyrics and a rhythm pattern selected from a plurality of types of rhythm patterns according to the number of notes. Techniques for generating are disclosed.

Fukayama et al., “Orpheus: Automatic Music System Based on Prosody of Lyrics”, Information Processing Society of Japan [Music Information Science], 2008 (78), p.179-184, July 30, 2008

  However, in the technique of Non-Patent Document 1, only a note sequence corresponding to the prosody of the lyrics is generated, and the relationship between each unit section (for example, a two-measure section) having a musical group in the music and the lyrics. Is not added at all. Therefore, for example, one phrase of the lyrics may straddle two unit sections of the music, and there is a possibility that a music that is unnatural and difficult to listen to is generated. In view of the above circumstances, an object of the present invention is to generate a song in which a character string such as lyrics appropriately corresponds to each unit section of the song.

  Means employed by the present invention to solve the above problems will be described. In order to facilitate understanding of the present invention, in the following description, the correspondence between each element of the present invention and the element of each of the embodiments described later is indicated in parentheses, but the scope of the present invention is not limited to the embodiment. It is not intended to limit the example.

  The music generation device of the present invention is a device that generates a music with melody assigned to each of a plurality of unit sections, and stores a plurality of melody data that specify different melody (for example, a storage device 24). A character string acquisition unit (for example, the information acquisition unit 32) for acquiring a designated character string composed of a plurality of characters, and a character allocation unit (for example, for distributing a plurality of character groups obtained by dividing the designated character string to each unit section A melody evaluation value (for example, a melody evaluation value (for example, the character allocation process S11) and the number of characters of each character group corresponding to the unit section and the number of melody notes specified by the melody data for each of the plurality of unit sections. Character string evaluation means (for example, music generation unit 34 / character string evaluation process S14) for calculating the melody evaluation value EB) for each melody data, and each melody data in the unit section for each of the plurality of unit sections. ; And a melody selecting means for selecting one of a plurality of melody data (e.g. music generating unit 34 / melody selecting process S15) in accordance with the melody evaluation value. In the above configuration, a plurality of character groups obtained by dividing the designated character string are distributed to each unit section. Therefore, it is possible to generate a musical piece in which each character group of the designated character string and the melody of each unit section naturally correspond to each other as compared with the technique of Non-Patent Document 1 that generates a note string according to the prosody of the lyrics. It is. Note that “melody” in the present invention means a time series (sequence of sounds) of a plurality of sounds.

  In a preferred embodiment of the present invention, each of the plurality of melody data includes section data (for example, section data DM2) that specifies boundary points that divide the melody specified by the melody data into a plurality of melody sections, and character string evaluation For each of the plurality of unit sections, the means associates each character group corresponding to the unit section with each melody section (for example, melody section B) of the melody data, and calculates a melody evaluation value according to the number of characters in each melody section. Calculation is performed for each melody data (for example, a second adjustment process SC12 and a third adjustment process SC13 described later). In the above aspect, since the melody index value corresponding to the number of characters corresponding to each melody section is calculated for each melody data, for example, the possibility of generating a song having too many characters corresponding to each melody section is reduced. It is possible.

  For example, for each of a plurality of unit sections, the character string evaluation means associates each character group corresponding to the unit section with each melody section of the melody data, and corresponds to the number of notes in each melody section and the melody section. The melody evaluation value of each melody data is calculated so that the melody evaluation value becomes a smaller numerical value as the difference from the number of characters is larger (for example, a second adjustment process SC12 described later). In the above aspect, the greater the difference between the number of notes in each melody section and the number of characters corresponding to the melody section, the smaller the melody evaluation value (that is, the lower the possibility of selection by the melody selection means). Thus, the possibility that a song having extremely different numbers of characters and notes in each melody section is generated can be reduced.

  Further, the character string evaluation means associates each character group corresponding to the unit section with each melody section of the melody data for each of the plurality of unit sections, and the number of characters corresponding to each melody section corresponds to the melody section. The melody evaluation value of each melody data is calculated so that the melody evaluation value becomes a small value when the allowable value according to the time length is exceeded (for example, a third adjustment process SC13 described later). For example, the allowable value is set to a smaller numerical value as the time length of the melody section is shorter. In the above embodiment, when the number of characters in each melody section exceeds the allowable value, the melody evaluation value is set to a small numerical value, so that the number of characters in each melody section is too large (that is, each character is uttered very quickly) The possibility that an unnatural music that needs to be generated) can be reduced.

  In a preferred aspect of the present invention, for each of the plurality of unit sections, the character string evaluation means increases the difference between the number of character groups corresponding to the unit section and the number of melody sections specified by the melody data. The melody evaluation value of each melody data is calculated so that the melody evaluation value becomes a small numerical value (for example, first adjustment process SC11 described later). In the above embodiment, the larger the difference between the number of character groups corresponding to a unit section and the number of melody sections, the smaller the melody evaluation value is set, so the number of character groups and the number of melody sections in the unit section. It is possible to reduce the possibility of generating music that is extremely different from.

  In a preferred aspect of the present invention, the character string evaluation means has a small melody evaluation value for each of the plurality of unit sections when the number of characters corresponding to the unit section exceeds the number of melody notes in the melody data. Then, a melody evaluation value of each melody data is calculated (for example, a fourth adjustment process SC14 described later). In the above embodiment, when the number of characters corresponding to each unit section exceeds the number of notes, the melodic evaluation value is set to a small value, so that melodic data whose number of notes is significantly smaller than the number of characters in each unit section can be selected. Can be reduced. Therefore, for example, in the configuration in which each note specified by the melody data is divided so that the note and the character correspond to each other, there is an advantage that the frequency of the division of the note is reduced and a music having the characteristics of the melody data can be generated.

  The music generation device according to a preferred aspect of the present invention includes a chord acquisition unit (for example, an information acquisition unit 32) that acquires chord progression data that specifies chord progression of a song, and chord progression data for each of a plurality of unit sections. Melody evaluation means (for example, music generation unit 34 / melody) for calculating for each melody data a melody evaluation value (for example, melody evaluation value EA) corresponding to the relationship between the chord progression in the specified unit section and the melody specified by the melody data. The character string evaluation means adjusts the melody evaluation value of each melody data calculated by the melody evaluation means. In the above mode, the melody evaluation value corresponding to the relationship between the chord progression in each unit section specified by the chord progression data and the melody specified by the melody data is calculated for each melody data and adjusted by the character string evaluation means. Is done. Therefore, as compared with the configuration of Non-Patent Document 1 in which the musical note string is restricted by the prosody of the lyrics, there is an advantage that various musical pieces corresponding to the chord progression specified by the chord progression data can be generated.

  The music generation device according to a preferred aspect of the present invention adjusts the number of melody notes specified by the melody data selected by the melody selection means for each unit section, and each melody note after adjustment and each of the notes in the unit section. It includes melody adjustment means (for example, music generation unit 34 / melody adjustment processing S16) for associating characters. In the above aspect, since the number of notes in the melody data is adjusted so that the notes and characters correspond to each other, it is possible to generate a natural musical piece in which the notes and characters clearly correspond.

  In a preferred aspect of the present invention, the melody adjustment means continues when the number of melody notes in the melody data selected by the melody selection means for each unit section is less than the number of characters corresponding to the unit section. The process of dividing one note having the longest length is repeated until the number of notes matches the number of characters in the unit section (for example, step SD12). In the above aspect, the number of notes is adjusted to be the same as the number of characters by dividing a note having a long duration, so that, for example, the melody data is compared with a configuration in which a selected note is divided regardless of the duration. There is an advantage that it is possible to generate natural music that maintains its characteristics.

  In a preferred aspect of the present invention, the melody adjusting means includes a plurality of melodic notes when the number of melody notes in the melody data selected by the melody selection means for each unit section exceeds the number of characters corresponding to the unit section. Among them, the same number of notes as the number of characters corresponding to the unit section are selected, and a long sound symbol is made to correspond to a non-selected note (for example, step SD14 and step SD15). In the above aspect, the same number of notes as the number of characters corresponding to the unit section are selected from the plurality of notes of the melody, and a long sound symbol is given to the non-selected notes. Therefore, there is an advantage that a natural music that maintains the characteristics of the melody data can be generated.

  Although the method by which the melody adjustment means selects the same number of notes as the number of characters corresponding to the unit section among the plurality of notes is arbitrary, for example, chord acquisition means (information) for acquiring chord progression data specifying the chord progression of the music The acquisition unit 32), a first evaluation value corresponding to the relationship between the chord specified by the chord progression data and the pitch of the note, and the continuation of the note for each of a plurality of notes of the melody specified by each melody data In a configuration including melody evaluation means (for example, music generation unit 34 / melody evaluation processing S13) for calculating a note evaluation value from the second evaluation value corresponding to the length, the melody data selected by the melody selection means for each unit section It is possible to select the same number of notes as the number of characters corresponding to the unit section in the order of the note evaluation values from a plurality of notes of the melody. In the above aspect, musically important notes are selected from a plurality of notes specified by the melody data, so that the effect of generating a natural musical piece maintaining the musical characteristics of the melody data is particularly remarkable. is there.

  The music generating apparatus according to a preferred aspect of the present invention is a voice synthesizing unit (for example, a voice synthesizing unit 44) that generates a voice signal of a voice in which the melody specified by the melody data adjusted by the melody adjusting unit is pronounced by a designated character string. And a tone generation unit (for example, a tone generation unit 42) that generates a tone signal of the performance sound of the melody specified by the tune data adjusted by the melody adjustment unit, and a voice signal and a sound generation unit generated by the speech synthesis unit Mixing means (for example, a mixing unit 46) for mixing the musical tone signals.

  The music generating device according to each of the above aspects is realized by hardware (electronic circuit) such as a dedicated DSP (Digital Signal Processor), or a general-purpose arithmetic processing device such as a CPU (Central Processing Unit) and a program. It is also realized through collaboration. The program according to the present invention includes a plurality of characters stored in a computer having storage means for storing a plurality of melody data for designating different melody in order to generate a melody with melody assigned to each of a plurality of unit sections. A character string acquisition process for acquiring a specified character string composed of: a character allocation process for distributing a plurality of character groups into which a specified character string is divided (for example, a character allocation process S11); For each, a character string evaluation process (for example, a character string evaluation process S14) for calculating for each melodic data a melodic evaluation value corresponding to the number of characters in each character group corresponding to the unit section and the number of melodic notes specified by the melodic data And, for each of the plurality of unit sections, a melody selection process (for example, a melody selection process) that selects any one of the plurality of melody data according to the melody evaluation value of each melody data in the unit section. S15) is executed. According to the above program, the same operation and effect as the music generating device according to the present invention are realized. Note that the program of the present invention is provided in a form stored in a computer-readable recording medium and installed in the computer, or is provided in a form distributed via a communication network and installed in the computer.

It is a block diagram of the music production | generation apparatus which concerns on 1st Embodiment of this invention. It is explanatory drawing of the calculation of the content of a melody data, and a melody evaluation value. It is explanatory drawing of each unit area which comprises a music, and each character group which comprises a designated character string. It is a schematic diagram of chord progression data. It is a flowchart of a music production | generation process. It is a flowchart of a character allocation process. It is a flowchart of a melody evaluation process. It is a flowchart of a character string evaluation process. It is a flowchart of a melody adjustment process. It is a schematic diagram of the unit data of chord progression data in the second embodiment. It is a schematic diagram of the designated character string in 3rd Embodiment. It is a schematic diagram of the designated character string in 6th Embodiment. It is a block diagram of the music production | generation apparatus which concerns on 6th Embodiment.

<First Embodiment>
FIG. 1 is a block diagram of a music generation device 100A according to the first embodiment of the present invention. The music generation device 100A is a signal processing device (automatic music device) that generates a music and outputs an acoustic signal V of the performance sound of the music. In the following description, a case where the song sound generates a song for which the melody is in charge is illustrated.

  An input device 12 and a sound emitting device 14 are connected to the music generating device 100A. The input device 12 is a device (for example, a mouse or a keyboard) that receives an instruction from a user. The sound emitting device 14 (for example, a speaker or a headphone) emits a sound wave corresponding to the acoustic signal V supplied from the music generating device 100A. In addition, illustration of the D / A converter etc. which convert the acoustic signal V from digital to analog is abbreviate | omitted for convenience.

  The music generation device 100 </ b> A of the first embodiment is realized by a computer system that includes an arithmetic processing device 22 and a storage device 24. The arithmetic processing device 22 executes a program PGM stored in the storage device 24 to thereby generate a plurality of functions (information acquisition unit 32, music generation unit 34, musical sound generation unit 42, voice synthesis unit 44). , The mixing unit 46) is realized. A configuration in which each function of the arithmetic processing unit 22 is distributed over a plurality of integrated circuits, or a configuration in which a dedicated electronic circuit (DSP) shares a part of the functions may be employed.

  The storage device 24 stores a program PGM stored in the arithmetic processing device 22 and various data used by the arithmetic processing device 22. A known recording medium such as a semiconductor recording medium or a magnetic recording medium can be arbitrarily employed as the storage device 24. It is also possible to configure the storage device 24 with a plurality of recording media. As shown in FIG. 1, the storage device 24 stores a plurality of melody data DM. Each of the plurality of melody data DM is data for specifying a melody (hereinafter referred to as “material melody”) used as a material for generating music. As shown in FIG. 1, note string data DM1 and section data DM2 are included. Consists of including.

  As shown in FIG. 2, the note string data DM1 designates a time series of a plurality of notes constituting a material melody. Specifically, the note string data DM1 designates a pitch and a pronunciation period (sound generation time and duration) for each of a plurality of notes of the material melody. Each note string data DM1 is generated using a melody of a predetermined length extracted from an existing musical piece as a material melody. The first embodiment exemplifies a case where a material melody is composed of two measures (16 eighth notes) of a 4/4 time tune. The note string data DM1 can be described in a format conforming to, for example, the MIDI (Musical Instrument Digital Interface) standard. Specifically, event data that specifies the pitch (note number) of each note in the material melody and specifies sounding and muting, timing data that specifies the processing interval of each event data (note duration and interval), Is suitable as the note string data DM1.

  The music generation device 100A of the first embodiment generates music composed of a plurality (K) of unit sections U as shown in FIG. One unit section U is a section corresponding to, for example, a predetermined number of measures of a music piece. In the following description, it is assumed that one unit section U corresponds to two bars (16 eighth notes) of a 4/4 time tune, similarly to the material melody. One of the plurality of melody data DM is sequentially selected for each of the K unit intervals U, and a musical piece in which the melody for each unit interval U is selected according to the melody data DM (note string data DM1) is generated.

  As shown in FIG. 2, the section data DM2 is data for dividing the material melody specified by the note string data DM1 into a plurality of sections (hereinafter referred to as “melody section”) B on the time axis. The boundary point of each melody section B is designated. For example, each melody section B is demarcated with a section having a musical grouping in the material melody as a unit. The section data DM2 is arbitrarily set in consideration of a musical impression after the producer of the melody data DM (provider of the music generation device 100A) listens to the material melody.

  Each melody data DM described above can be described as a music file in SMF (Standard MIDI File) format, for example. Specifically, the note string data DM1 and the section data DM2 are described as separate tracks (channels). The section data DM2 is described as, for example, data in MIDI format in which event data meaning boundary points and timing data specifying the time points of each event data (time points of boundary points) are arranged.

  The information acquisition unit 32 in FIG. 1 acquires information applied to music generation. The information acquisition unit 32 of the first embodiment generates the chord progression data DC, the designated character string DL, and the variable data DX in accordance with an instruction from the user to the input device 12. As shown in FIG. 1, information (DC, DL, DX) generated by the information acquisition unit 32 is stored in the storage device 24. A configuration in which the information acquisition unit 32 acquires information (DC, DL, DX) from a portable or stationary recording medium, or a configuration in which the information acquisition unit 32 acquires information from another communication terminal via a communication network. Can also be employed.

  The chord progression data DC designates the chord progression of the music. Specifically, the chord progression data DC has a structure in which unit data DU corresponding to each beat point (for example, every eighth note) in the music is arranged in time series as shown in FIG. Specifically, each unit data DU designates a bar number, a beat number, and a code. For example, the unit data DU written as “1-1: C” in FIG. 4 designates “C” as the chord of the first beat (first eighth note) of the first measure of the music, and “1” The unit data DU written as “−5: F” designates “F” as the code of the fifth beat of the first measure of the music. Unit data DU that does not include a code designation (for example, unit data DU represented as “1-2:” in FIG. 4) means that the code designated immediately before is maintained.

  The chord progression data DC is described as a text file, for example. That is, as illustrated in FIG. 4, the user sequentially inputs the contents of the unit data DU for each line from the input device 12, and the information acquisition unit 32 converts the document file created by the user into the code progression data. It is stored in the storage device 24 as DC. The user can arbitrarily set the total number of unit data DU in the chord progression data DC. The total number K of unit sections U constituting the music (the time length of the music) is variably set according to the total number of unit data DU in the chord progression data DC. In the first embodiment, one unit data DU corresponds to an eighth note, and one unit section U corresponds to two bars of 4/4 time (16 eighth notes). When the chord progression data DC includes 64 unit data DU, a song composed of 4 unit sections U (8 bars corresponding to 64 eighth notes) is generated.

  The designated character string DL is a character string of the lyrics of the music. The information acquisition unit 32 variably sets the designated character string DL in accordance with an instruction (character input) from the user to the input device 12. Variable data DX designates a variable applied to music generation. The information acquisition unit 32 variably sets the numerical value of each variable according to an instruction from the user to the input device 12. Specific examples of variables designated by the variable data DX will be described later. As can be understood from the above description, the information acquisition unit 32 realized by the arithmetic processing device 22 includes an element for acquiring the chord progression data DC (code acquisition means) and an element for acquiring the designated character string DL (character string acquisition means). ), And functions as an element for obtaining the variable data DX.

  1 generates music data Q by using each melody data DM, chord progression data DC, designated character string DL, and variable data DX. The music data Q is data for specifying the melody of the music, and is composed of time series of melody data QU corresponding to each unit section U as shown in FIG. FIG. 5 is a flowchart of a process in which the music generation unit 34 generates the music data Q (hereinafter referred to as “music generation process”). When the chord progression data DC, the designated character string DL, and the variable data DX are stored in the storage device 24 and the user instructs the start of music generation by operating the input device 12, the music generation processing of FIG. Be started.

  When the music generation process is started, the music generation unit 34 executes a character allocation process S11 for associating the designated character string DL stored in the storage device 24 with each of the K unit intervals U. FIG. 6 is a flowchart of the character assignment process S11. When the character assignment process S11 is started, the music generation unit 34 converts the designated character string DL into hiragana (SA11), and then divides the designated character string DL into a plurality of character groups G as shown in FIG. 3 (SA12). ). In the first embodiment, the designated character string DL is divided into character groups G for each phrase. Therefore, the number of characters differs for each character group G. Known natural language processing such as morphological analysis can be arbitrarily used to specify the phrase of the designated character string DL. If the designated character string DL is input as a hiragana character, the process of step SA11 is omitted.

  As shown in FIG. 3, the music generation unit 34 distributes a plurality of character groups G into which the designated character string DL is divided into K unit sections U (SA13). Specifically, the time series of the plurality of character groups G of the designated character string DL is divided into K sets corresponding to different unit sections U. For example, the number of character groups G distributed to each unit section U is made as uniform as possible for the K unit sections U (the difference in the number of character groups G for each unit section U is minimized). A plurality of character groups G are divided into unit sections U in chronological order. As described above, when the arithmetic processing unit 22 executes the character allocation process S11, an element (character allocation unit) that distributes the plurality of character groups G of the designated character string DL to each unit section U is realized. .

  When the character assignment process S11 ends, the music generation unit 34 selects one unit section U (hereinafter referred to as “target unit section U”) from the K unit sections U (S12). In step S12 immediately after the start of the music generation process, the first unit section U is selected as the target unit section U. The music generation unit 34 executes a melody generation process (S13 to S16) for generating melody data QU for the target unit section U, and determines whether or not the melody generation process has been completed for each of the K unit sections U. (S17). When there is a unit section U for which melody data QU has not been generated (S17: NO), the music generation unit 34 selects a unit section U immediately after the current target unit section U as a new target unit section U. (S12), the melody generation process (S13 to S16) is executed for the updated target unit section U. That is, the melody generation process (S13 to S16) is sequentially repeated for each of the K unit intervals U, so that the melody data QU of each unit interval U of the music is generated. In the first embodiment, the K unit intervals U are sequentially selected in the order from the beginning to the end. However, the K unit intervals U may be sequentially selected in the order from the end to the beginning. .

  The melody generation process (S13 to S16) includes a melody evaluation process S13, a character string evaluation process S14, a melody selection process S15, and a melody adjustment process S16. The melody evaluation processing S13 stores a melody evaluation value EA corresponding to the relationship (musical correlation) between the chord progression designated by the chord progression data DC and the melody designated by the melody data DM for the target unit section U in the storage device 24. Is a process for calculating each of the plurality of melody data DM. In the character string evaluation process S14, a melody evaluation value EB corresponding to the relationship between each character group G distributed to the target unit section U in the character assignment process S11 and the melody specified by the melody data DM is stored in a plurality of melodies in the storage device 24. This is a process for calculating each of the data DM. In the first embodiment, the melody evaluation value EB is calculated by adjusting the melody evaluation value EA calculated in the melody evaluation processing S13 by the character string evaluation processing S14.

  The melody selection process S15 is a process of selecting one melody data DM from a plurality of melody data DM in the storage device 24 in accordance with the melody evaluation value EB calculated for each melody data DM. Specifically, melodic data DM having the maximum melodic evaluation value EB is selected from among a plurality of melodic data DM. The melody adjustment process S16 is a process for generating the melody data QU of the target unit section U by adjusting one melody data DM selected in the melody selection process S15. The melody generation process (S13 to S16) outlined above is sequentially executed for each of the K unit intervals U. Specific contents of each processing (S13 to S16) will be described below.

<Melody evaluation process S13>
FIG. 7 is a flowchart of the melody evaluation process S13. When the melody evaluation process S13 is started, the music generation unit 34 selects one melody data DM (hereinafter referred to as “target melody data DM”) among the plurality of melody data DM stored in the storage device 24 (SB11). . Then, the music generation unit 34 calculates a note evaluation value e for each of the plurality of notes of the material melody specified by the note string data DM1 of the target melody data DM (SB12).

  The note evaluation value e is calculated in accordance with the first evaluation value e1 and the second evaluation value e2, as shown in FIG. The music generation unit 34 of the first embodiment calculates a multiplication value of the first evaluation value e1 and the second evaluation value e2 as a note evaluation value e. The first evaluation value e1 of one note is set to a numerical value corresponding to the relationship between the chord ZC designated by the chord progression data DC and the pitch ZP of the note at the time of the note's pronunciation.

  Specifically, the first evaluation value e1 is set to the numerical value a1 when the pitch ZP of the note corresponds to the constituent sound of the chord ZC, and the pentatonic constituent sound (the pitch ZP is the root note of the chord ZC) ( (Except for the constituent sound of the chord ZC), the numerical value a2 is set. When the pitch ZP is a stem sound that does not correspond to either the constituent sound of the chord ZC or the constituent sound of the pentatonic, the numerical value a3 is set. Is set. If the note pitch ZP does not correspond to any of the notes, the first evaluation value e1 is set to zero.

  Each of the numerical values a1 to a3 is a positive number designated by the variable data DX. The information acquisition unit 32 individually sets each of the numerical values a1 to a3 in accordance with an instruction from the user to the input device 12. In FIG. 2, the numerical value a1 corresponding to the constituent sound of the chord is set to “5”, the numerical value a2 corresponding to the constituent sound of the pentatonic is set to “3”, and the numerical value a3 corresponding to other stem sounds is set to “1”. ”Is assumed. Accordingly, when attention is paid to the section in which the chord ZC designated by the chord progression data DC is “F”, as shown in FIG. 2, the musical note whose pitch ZP is set to “F (fa)” (the constituent sound of the chord). The first evaluation value e1 of the note is set to “5 (a1)”, and the first evaluation value e1 of the note having the pitch ZP set to “D (re)” (a pentatonic constituent sound) is “3 (a2)”. And the first evaluation value e1 of the note whose pitch ZP is set to “E (mi)” (stem tone not corresponding to any of them) is set to “1 (a3)”.

  On the other hand, the second evaluation value e2 is set to a numerical value (for example, the number of ticks in the sounding interval of each note) corresponding to the duration of each note. Specifically, as can be understood from FIG. 2, the second evaluation value e2 is set to a larger value as the duration of the note is longer. The multiplication value of the first evaluation value e1 and the second evaluation value e2 described above is calculated as the note evaluation value e for each of the plurality of notes specified by the target melody data DM.

  The music generation unit 34 calculates the melody evaluation value EA of the target melody data DM from the note evaluation value e of each note designated by the target melody data DM (SB13, SB14). Specifically, the music generation unit 34 calculates the total value Ta of the note evaluation values e over all the notes specified by the target melody data DM (SB13), and the target melody data DM is the total of the durations of all the notes. The melody evaluation value EA (EA = Ta / Tb) is calculated by dividing the total value Ta of the note evaluation value e by the value (total value of the second evaluation value e2) Tb (SB14). The division by the total value Tb is an operation for normalizing to a numerical value that compensates for the difference in the total value Tb for each material melody.

  As can be understood from the above description, assuming that the numerical value a1 and the numerical value a2 are set to a large numerical value compared to the numerical value a3, the chord component sound and the pentatonic component sound specified by the chord progression data DC There is a tendency that the melody evaluation value EA becomes a larger numerical value (that is, a numerical value easy to be selected in the melody selection process S15) as the material melody has a longer duration. In other words, the melody evaluation value EA can be used as an index of the degree to which the chord specified by the chord progression data DC and each note of the material melody are musically correlated (harmonized).

  When the melody evaluation value EA of the target melody data DM is calculated in the above procedure, the music generation unit 34 determines whether or not the calculation of the melody evaluation value EA has been completed for all the melody data DM stored in the storage device 24. (SB15). When there is uncalculated melody data DM of the melody evaluation value EA (SB15: NO), the music generation unit 34 selects the melody data DM for which the melody evaluation value EA has not been calculated as new target melody data DM ( SB11), the melodic evaluation value EA is calculated for the updated target melody data DM (SB12 to SB14). On the other hand, the melody evaluation process S13 ends when the calculation of the melody evaluation value EA is completed for all the melody data DM (SB15: YES). As understood from the above description, the arithmetic processing unit 22 executes the melody evaluation process S13, thereby realizing an element (melody evaluation means) for calculating the melody evaluation value EA for each melody data DM (for each material melody). Is done.

<Character string evaluation process S14>
FIG. 8 is a flowchart of the character string evaluation process S14. When the character string evaluation process S14 is started, the music generation unit 34 selects one target melody data DM from the plurality of melody data DM stored in the storage device 24 (SC10). The music generation unit 34 sequentially adjusts the melody evaluation value EA calculated for the target melody data DM in the melody evaluation process S13 by each process from the first adjustment process SC11 to the fourth adjustment process SC14 described in detail below. Calculate the melodic evaluation value EB.

In the first adjustment process SC11, the number NA of the character group G distributed to the target unit section U in the character assignment process S11 and the number MA of the melody section B obtained by dividing the material melody according to the section data DM2 of the target melody data DM. This is a process for adjusting the melody evaluation value EA of the target melody data DM in accordance with the difference (absolute value) ΔA. Specifically, the melody evaluation value EA is adjusted to a smaller numerical value as the difference ΔA is larger. For example, the melody evaluation value EA is updated to a value obtained by multiplying a predetermined numerical value b1 to the power of ΔA by the current melody evaluation value EA (EA = EA × b1 ^ΔA ). The numerical value b1 is a positive number of 1 or less. As can be understood from the above description, the melody data DM of the material melody having a large difference ΔA between the number NA of the character group G and the number MA of the melody section B tends to be difficult to select in the melody selection process S15. That is, a song in which the number NA of character groups G and the number MA of melody intervals B are extremely different (for example, a song in which an excessively large number of character groups G correspond to one melody interval B, or an excessively large number of melody intervals) The possibility that a music piece corresponding to one character group G) is generated is reduced.

The second adjustment process SC12 associates each character group G assigned to the target unit section U with each melody section B of the material melody specified by the target melody data DM, and the number of characters in the character group G in each melody section B. This is a process for adjusting the melody evaluation value EA according to the difference (absolute value) ΔB between NB and the number of notes MB in the melody section B. Specifically, first, the music generation unit 34 mutually sets each melody section B of the material melody specified by the target melody data DM and each character group G assigned to the target unit section U in time series order. The difference ΔB is calculated for all combinations to be matched. The difference ΔB is a numerical value obtained by adding the difference between the number of characters NB and the number of notes MB of the character group G in the melody section B for all the melody sections B of the material melody. Secondly, the music generation unit 34 determines the correspondence between each melody section B and each character group G so that the difference ΔB is minimized, and thirdly, the larger the difference ΔB in the relationship after the determination, the higher the melody evaluation. Adjust the value EA to a smaller number. For example, the melody evaluation value EA is updated to a value obtained by multiplying a predetermined numerical value b2 to the power of ΔB by the melody evaluation value EA after execution of the first adjustment process SC11 (EA = EA × b2 ^ΔB ). The numerical value b2 is a positive number of 1 or less. As understood from the above description, the melody data DM of the material melody having a large difference ΔB between the number of characters NB and the number of notes MB in each melody section B tends to be difficult to select in the melody selection process S15. is there. That is, in each melody section B, the number of characters NB and the number of notes MB are extremely different (for example, a song in which an excessively large number of characters correspond to a single melody interval B, or an excessively large number of notes. The possibility that a music piece corresponding to one character group G) is generated is reduced.

The third adjustment process SC13 is a process for reducing the melodic evaluation value EA when the number of characters NB corresponding to each melodic section B exceeds a predetermined allowable value TH. The allowable value TH is set to the number of sixteenth notes that can be included in each melody section B (therefore, a variable value corresponding to the time length of the melody section B). Since the time length of each melody section B can be different, the allowable value TH is also calculated for each melody section B individually. The music generation unit 34 calculates the difference ΔC between the allowable value TH and the number of characters NB for one melody section B (ΔC = TH−ZC), and when the difference ΔC is a positive number (that is, one character) Melody evaluation value EA is maintained when a time length longer than a sixteenth note is secured in melody section B), while a difference ΔC is a negative number (a sixteenth note time length for one character) When the melody section B cannot be secured), the melody evaluation value EA is decreased. Specifically, the melody evaluation value EA is updated to a value obtained by multiplying the melody evaluation value EA after execution of the second adjustment process SC12 by the power of ΔC (negative number) of a predetermined numerical value b3 (EA = EA × b3 ^ΔC). ). The numerical value b3 is a positive number of 1 or more. The above process is repeated for each melody section B in the material melody. As understood from the above description, the melody data DM of the material melody in which the number NA of characters in each melody section B exceeds the allowable value TH tends to be difficult to select in the melody selection process S15. That is, the possibility that a song with an excessively fast utterance speed of each character (for example, a song with each character uttered at a speed exceeding the 16th note) is reduced.

  In the fourth adjustment process SC14, the number of characters corresponding to the target unit section U (the total number of characters of each character group G distributed to the target unit section U) ND is the number of notes constituting the material melody of the target melody data DM. (Total value of the number of notes in each melody section B of the material melody) This is a process for reducing the melody evaluation value EA when it exceeds MD. Specifically, the music generation unit 34 maintains the melody evaluation value EA when the number of characters ND in the target unit section U is equal to or less than the number MD of material melody notes, and the characters in the target unit section U are maintained. If the number ND exceeds the number MD of material melody notes MD, the melody evaluation value EA after execution of the third adjustment process SC13 is multiplied by a predetermined numerical value b4 (EA = EA × b4). The numerical value b4 is a positive number of 1 or less. As understood from the above description, the melody data DM of the material melody in which the number of characters ND corresponding to the target unit section U exceeds the number of notes MD tends to be difficult to select in the melody selection process S15. That is, the possibility that the number of characters in the unit section U is excessively larger than the number of notes is reduced.

  Each of the numerical value b1 to the numerical value b4 applied to the character string evaluation process S14 is a variable designated by the variable data DX in the same manner as the numerical value a1 to the numerical value a3 in the first embodiment, and according to an instruction from the user. Set to variable.

  When the melody evaluation value EB of the target melody data DM is calculated by sequentially executing the processes (SC11 to SC14) described above on the melody evaluation value EA of the target melody data DM, the music generation unit 34 stores the storage device 34. It is determined whether or not the calculation of the melody evaluation value EB has been completed for all the melody data DM stored in 24 (SC15). When there is melody data DM for which the melody evaluation value EB has not been calculated (SC15: NO), the music generation unit 34 selects the melody data DM for which the melody evaluation value EB has not been calculated as new target melody data DM. (SC10) A melody evaluation value EB is calculated for the updated target melody data DM (SC11 to SC14). On the other hand, the character string evaluation process S14 ends when the calculation of the melody evaluation value EB is completed for all the melody data DM (SC15: YES). As can be understood from the above description, the arithmetic processing unit 22 executes the character string evaluation process S14, whereby the number of characters or the number of character groups G corresponding to each unit section U and the melody data DM specify the material melody. An element (character string evaluation means) for calculating a melody evaluation value EB corresponding to the number of notes or the number of melody sections B for each melody data DM is realized. Note that the order of the first adjustment process SC11 to the fourth adjustment process SC14 is changed as appropriate.

<Melody selection process S15>
The music generation unit 34 stores a plurality of pieces of melody data DM (hereinafter referred to as “selected melody data DM”) having the maximum melody evaluation value EB calculated for each melody data DM in the character string evaluation process S 14 in the storage device 24. Is selected from the melody data DM (S15). That is, when the arithmetic processing unit 22 executes the melody selection process S15, an element (melody selection means) for selecting the melody data DM according to the melody evaluation value EB is realized. Note that the method in which the music generation unit 34 selects the selected melody data DM in accordance with the melody evaluation value EB of each melody data DM is appropriately changed. For example, a method in which the music generation unit 34 randomly selects one selected melody data DM from a plurality of melody data DM in which the melody evaluation value EB exceeds a predetermined threshold, or a predetermined position that is higher in the descending order of the melody evaluation value EB. A method in which the music generation unit 34 randomly selects one piece of selected melody data DM from one piece of melody data DM may be employed. According to the above method, it is possible to generate a separate musical piece every time a musical piece is generated.

<Melody adjustment processing S16>
The music generation unit 34 adjusts the number of notes specified by the selected melody data DM selected in the melody selection process S15 so that characters and notes correspond one-to-one in the target unit section U. The melody data QU for the section U is generated (S16). FIG. 9 is a flowchart of the melody adjustment process S16.

  When the melody adjustment process S16 is started, the music generation unit 34 determines whether or not the number of characters ND distributed to the target unit section U exceeds the number of notes MD specified by the selected melody data DM (SD11). . When the number of characters ND exceeds the number of notes MD (SD11: YES), the music generation unit 34 increases the number of notes MD to the number of characters ND (SD12). Specifically, the music generation unit 34 performs a process of dividing one note having the longest duration among a plurality of notes designated by the selected melody data DM into two notes, and the number of notes MD is a character. Iteratively repeats until the number ND is reached. One note is divided at a time corresponding to a beat point, for example. The selected melody data DM after the processing of step SD12 is determined as the melody data QU of the target unit section U.

  On the other hand, if the determination result in step SD11 is negative (SD11: NO), the music generation unit 34 determines whether the number of characters ND is less than the number of notes MD (SD13). When the number of characters ND is less than the number of notes MD (SD13: YES), the music generation unit 34 selects ND notes of the same number as the characters from the MD notes designated by the selected melody data DM (SD14). . For the selection of ND notes, the note evaluation value e calculated for each note in the melody evaluation process S13 (SB12) is used. Specifically, the music generation unit 34 selects ND notes positioned higher in the descending order (large → small) of the note evaluation value e among the MD notes designated by the selected melody data DM. However, the first note (ie, musically important note) in the melody section B is preferentially selected regardless of the note evaluation value e. In addition, the music generation unit 34 processes notes not selected in step SD14 among the MD notes designated by the selected melody data DM (SD15). For example, the music generation unit 34 deletes a non-selected note and extends the immediately preceding note in the time axis direction by the deleted amount. However, if the pitch difference between the unselected note and the immediately preceding note falls below a threshold (for example, about 4 semitones), a long clef symbol (-) that means the continuation of the character of the immediately preceding note is changed to the unselected note. Make it correspond. Adding a long clef only to notes whose pitch is close to the immediately preceding note is that the music will be unnatural if the character continues from the immediately preceding note, even for notes whose pitch changes significantly from the immediately preceding note. Because it is perceived as an impression. The selected melody data DM after the processing of step SD15 is determined as the melody data QU of the target unit section U.

  On the other hand, when the determination result of step SD13 is negative (ie, when the number of notes MD matches the number of characters ND), the music generation unit 34 adjusts the number of notes MD (SD12, SD14, SD15). Do not execute. That is, the selected melody data DM is determined as the melody data QU of the target unit section U. By the processing (SD11 to SD15) described above, the number of characters ND and the number of notes MD match. That is, melody data QU is generated in which each character of the designated character string DL and each note of the material melody correspond one-to-one. As can be understood from the above description, the arithmetic processing unit 22 executes the melody adjustment process S16, thereby adjusting the number MD of material melody notes designated by the selected melody data DM of the target unit section U to each character. The element (melody adjustment means) corresponding to is realized.

  In the melody adjustment process S16 exemplified above, characters and notes are made to correspond one-to-one in the target unit section U, but the melody section B between the melody section B and the character group G corresponding to each other is used. It is also possible to associate each note in the character group with each character in the character group G on a one-to-one basis. For example, as in the second adjustment process SC12 described above, the music generation unit 34 makes each melody such that the difference ΔB between the number MB of notes in the melody section B and the number NB of characters in the character group G is minimized. The correspondence between the section B and each character group G is determined, and for each combination of the melody section B and the character group G, the number of notes in the melody section B and the number of characters in the character group G match. Processing S16 (SD11 to SD15) is executed.

  The specific contents of the melody generation process (S13 to S16) are as described above. As described above, the melody generation process is repeated for each of the K unit sections U, thereby generating K pieces of melody data QU constituting the music data Q. The music data Q generated by the music generator 34 is supplied to the musical sound generator 42 and the voice synthesizer 44 shown in FIG.

  The musical sound generation unit 42 generates a musical sound signal V1 indicating the performance sound of each note designated by the music data Q. Specifically, a well-known MIDI sound source is preferably employed as the musical sound generation unit 42. On the other hand, the voice synthesizer 44 generates a voice signal V2 corresponding to the music data Q generated by the music generator 34 and the designated character string DL generated by the information acquisition unit 32. The audio signal V2 is an acoustic signal indicating a singing sound uttered by each pitch of the designated character string DL at a pitch specified by the music data Q as a note corresponding to the character. For the generation of the audio signal V2, for example, a well-known unit connection type speech synthesis process can be suitably employed. That is, the speech synthesizer 44 sequentially selects speech segments corresponding to each character of the designated character string DL, and adjusts each speech segment to the pitch and pronunciation period of each note designated by the music data Q. Then, the audio signal V2 is generated by connecting them together.

  The mixing unit 46 generates the acoustic signal V by mixing (weighted sum) of the musical sound signal V1 generated by the musical sound generating unit 42 and the audio signal V2 generated by the voice synthesizing unit 44. The acoustic signal V generated by the mixing unit 46 is supplied to the sound emitting device 14 to reproduce the sound wave.

  As described above, in the first embodiment, the plurality of character groups G obtained by dividing the designated character string DL are distributed to each unit section U. Therefore, as compared with the technique of Non-Patent Document 1 that generates a note string according to the prosody of the lyrics, a natural song is generated in which each character group G of the designated character string DL and each unit section of the song correspond reasonably. Is possible. For example, if the designated character string DL is divided into a plurality of character groups G in phrase units as illustrated in the first embodiment, the musical grouping of music and the semantic grouping of the designated character string DL correspond to each other. It is possible to generate natural music. That is, for example, the possibility that an unnatural piece of music in which the boundary of musical groupings is located in the middle of a phrase is reduced.

  In the first embodiment, the melody evaluation value EB corresponding to the relationship between the number of characters or character groups G corresponding to each unit section U and the number of notes or melody sections B specified by the melody data DM is obtained from the melody data DM. It is calculated every time and applied to the selection of the melody data DM. Therefore, for example, the number of characters or character groups G and the number of notes or melody sections B are compared with the configuration in which the melody data DM of each unit section U is selected according to the melody evaluation value EA calculated in the melody evaluation processing S13. There is an advantage that it is possible to reduce the possibility of generating an unnatural music that is different from the above.

  By the way, in the technique of Non-Patent Document 1, since the musical note string is constrained so as to correspond to the prosody of the lyrics, there is a possibility that monotonous music deviating from the intention of the user is generated. In the first embodiment, a melody evaluation value EA corresponding to chord progression data DC designating the chord progression of music is calculated for each melody data DM and reflected in selection of melody data DM. Therefore, compared with the technique of nonpatent literature 1, there exists an advantage that the natural music corresponding to a user's intention is produced | generated.

  In addition, the numerical values (a1 to a3) of variables applied to the calculation of the melodic evaluation value EA and the numerical values of variables (b1 to b4) applied to the calculation of the melodic evaluation value EB are variable according to instructions from the user. Is set. Therefore, there is an advantage that music suitable for the user's intention can be generated. For example, if the numerical value a1 and the numerical value a2 are compared with the numerical value a3 and set to a large numerical value, the musical composition in which the constituent sound of the chord specified by the chord progression data DC or the constituent sound of the pentatonic predominates (the chord progression data DC specifies If a numerical value a3 is set to a value larger than the numerical value a1 or the numerical value a2, a musical piece deviating from the chord progression specified by the chord progression data DC is generated. Further, by appropriately adjusting the numerical values (b1 to b4) applied to the character string evaluation process S14, there is an advantage that a music piece in which characters and notes naturally correspond can be generated.

Second Embodiment
A second embodiment of the present invention will be described below. In addition, about the element which an effect | action and a function are equivalent to 1st Embodiment in each form illustrated below, each reference detailed in the above description is diverted and each detailed description is abbreviate | omitted suitably.

  FIG. 10 is a schematic diagram of one unit data DU constituting the chord progression data DC in the second embodiment. As shown in FIG. 10, in the unit data DU of the second embodiment, it is possible to specify the first specific sound, the second specific sound, and the deduction sound in addition to the specification of the code similar to the first embodiment. It is. In the example of FIG. 10, the pitch “D (re)” designated by the symbol “H:” means the first specific sound, and the pitch “A (ra)” designated by the symbol “L:” is This means the second specific sound, and the pitch “F (F)” designated by the symbol “Minus:” means the deduction sound. The first specific sound, the second specific sound, and the deduction sound are set according to an instruction from the user. Note that the effect of designating each note (first specific sound, second specific sound, deducted sound) in one unit data DU is effective only at the time (beat point) corresponding to the unit data DU. It is also possible to keep it valid until the designation is canceled in the unit data DU behind the time.

  In the melody evaluation process S13 (SB12), the music generation unit 34 generates a chord when the note specified by the target melody data DM corresponds to the first specific sound specified by the unit data DU at the time of pronunciation of the note. The first evaluation value e1 is set to the numerical value a4 regardless of whether or not it corresponds to the constituent sound or pentatonic constituent sound. Similarly, when the note corresponds to the second specific sound, the first evaluation value e1 is set to the numerical value a5, and when the note corresponds to the deduction sound, the first evaluation value e1 is set to the numerical value a6. Each of the numerical value a4 to the numerical value a6 is a variable designated by the variable data DX, similarly to the numerical value a1 to the numerical value a3 of the first embodiment, and is variably set according to an instruction from the user.

  The numerical values a4 and a5 are set to positive numbers (a4> a5). For example, if the numerical value a4 is set to a large numerical value from the numerical value a1 to the numerical value a3, the material melody including the note corresponding to the first specific sound can be preferentially selected in the melody selection processing S15. Further, if the numerical value a5 is set to a small value compared with the numerical value a1 or the numerical value a2, the possibility that a material melody having many notes corresponding to the second specific sound is selected in the melody selection processing S15 is reduced. On the other hand, the numerical value a6 is set to a negative number. Therefore, the note evaluation value e of the note corresponding to the deduction sound is a negative number (that is, a numerical value that lowers the melody evaluation value EA). That is, it is possible to avoid the selection of the material melody including the note corresponding to the deduction sound with high accuracy.

  In the second embodiment, the same effect as in the first embodiment is realized. In the second embodiment, the first specific sound, the second specific sound, and the deduction sound are designated by the chord progression data DC, and the numerical value (a4 to a6) of the first evaluation value e1 can be individually set for each. Therefore, there is an advantage that music that accurately reflects the user's intention can be generated. For example, it is possible to generate music that includes notes that the user wants to prioritize and music that does not include notes that the user wants to avoid.

<Third Embodiment>
In the first embodiment, the designated character string DL is divided into a plurality of character groups G in phrase units. In the third embodiment, the user can arbitrarily specify the boundary of the character group G. FIG. 11 is a schematic diagram of the designated character string DL input by the user by operating the input device 12. The designated character string DL is composed of a character string extending over a plurality of lines with a line feed symbol 82 as a boundary, and a delimiter symbol 84 can be inserted between adjacent characters. FIG. 11 illustrates the case where “: (colon)” is used as the delimiter symbol 84, but punctuation marks (punctuation marks and punctuation marks) may be used as the delimiter symbol 84. The user can arbitrarily select the position of the line feed symbol 82 and the delimiter symbol 84 of the designated character string DL along with the input of the character string.

  In the character assignment process S11 of FIG. 6, the music generation unit 34 divides the designated character string DL into a plurality of character groups G with the line feed symbol 82 and the delimiter symbol 84 as boundaries (SA12). Then, the music generation unit 34 distributes each character group G to K unit intervals U so that each character group G corresponds to a separate unit interval U before and after the line feed symbol 82 set in the designated character string DL. (SA13). When the line feed symbol 82 is not set in the designated character string DL, as in the first embodiment, a plurality of designated character strings DL are arranged so that the number of character groups G is equalized for K unit intervals U. Character group G is divided for each unit section U.

  In the third embodiment, the same effect as in the first embodiment is realized. In the third embodiment, since the user can arbitrarily designate the character group G of the designated character string DL, the natural character obtained by distributing the designated character string DL to each unit section U with the character group G designated by the user as a unit. It is possible to generate simple music. In addition to the position of the line feed symbol 82 and the delimiter symbol 84, the designated character string DL can be divided into character groups G at the boundaries of the clauses.

<Fourth embodiment>
The voice synthesizer 44 of the fourth embodiment executes a process (facial expression imparting process) for imparting a musical expression to the synthesized sound. A specific example of the facial expression imparting process is as follows.

(1) When the pitch of the note designated by the music data Q changes from the pitch ZP1 to the pitch ZP2, portamento (shakku) is added to the head of the note of the pitch ZP2. For example, if the pitch ZP2 exceeds the pitch ZP1, the effect of continuously increasing the pitch to the target pitch ZP2 after decreasing to a value below the pitch ZP1 (ascending portamento) Add to the beginning. Also, if the pitch ZP2 is lower than the pitch ZP1, the effect of continuously lowering the pitch to the target pitch ZP2 (descending portamento) after increasing to a value higher than the pitch ZP1 Add to the beginning.
(2) Add vibrato to each note. Specifically, for example, vibrato is added to the whole or a part of a note whose continuation length exceeds a predetermined threshold (for example, a section extending from the end point to a predetermined length ahead).
(3) When the number of characters per predetermined time exceeds the threshold (that is, when the synthesized sound is a quick mouth), the intelligibility of the synthesized sound (degree of opening at the time of utterance) is reduced, and the speech is made without opening the mouth completely Control the voice so that it is ambiguous.
(4) The volume of the first note of the music or the first note of each unit section U (or each melody section B) is increased. Further, instead of increasing the volume (or with increasing volume), a bar (fist) can be given.
(5) When the time length of the interval between adjacent notes exceeds a predetermined threshold, a breathing sound is added within the interval. The breathing voice is prepared in advance as a voice segment, for example. Further, when the time length of the interval between adjacent notes is below a predetermined threshold, each note is continued on the time axis. That is, legato is given between successive notes.
(6) A table in which pitches and rhythm patterns are associated with temporal changes in volume is prepared, and temporal changes in volume corresponding to the pitch patterns of notes in the melody section are given. In addition, a configuration in which the volume is temporarily increased immediately after the pitch of the note is changed, or a configuration in which the volume is changed according to the duration of the note (for example, the volume is increased as the duration is longer) may be employed.
(7) A time change of a control variable (for example, volume or intelligibility) for controlling the musical expression of the synthesized sound is set in advance in each melody data DM and applied during speech synthesis.
(8) A speech segment is prepared for each speech having different characteristics, and the speech segment actually applied to speech synthesis is changed according to the type of word included in the designated character string DL, chord progression, tone, etc. To do.

  The above is a specific example of the facial expression imparting process. In the fourth embodiment, the same effect as in the first embodiment is realized. Further, in the fourth embodiment, since the facial expression imparting process is executed when the voice synthesizer 44 executes voice synthesis (when the voice signal V2 is generated), it is possible to generate synthesized sounds with various musical expressions. There is.

<Fifth Embodiment>
Each melody data DM of the fifth embodiment includes attribute data DM3 and attribute data DM4 in addition to the note string data DM1 and the section data DM2. The attribute data DM3 designates at which position in the music the material melody is suitable. Specifically, the attribute data DM3 designates one of the beginning, center and end of the music. That is, the material melody attribute data DM3 suitable for the beginning of the music specifies the beginning, and the material melody attribute data DM3 suitable for the end of the music specifies the end. The attribute data DM4 designates whether or not the material melody is a particularly impressive melody. Specifically, the attribute data DM4 is set to the valid state when the material melody is an impressive melody, and is set to the invalid state when the melody is not impressive. The attribute data DM3 and the attribute data DM4 are arbitrarily set in consideration of a musical impression after the producer of the melody data DM listens to the material melody, for example.

  The music generation unit 34 executes the fifth adjustment process to the seventh adjustment process illustrated below in addition to the first adjustment process SC11 to the fourth adjustment process SC14 described above when the character string evaluation process S14 is executed. The fifth adjustment process to the seventh adjustment process are processes for adjusting the melody evaluation value EA calculated for each melody data DM in the melody evaluation process S13.

  The fifth adjustment process is a process of adjusting the melody evaluation value EA according to the attribute data DM3 of the target melody data DM. That is, the music generation unit 34 maintains the melody evaluation value EA when the position on the time axis of the target unit section U (the head / center / end in the music) matches the position specified by the attribute data DM3. If the position of the target unit section U does not match the position specified by the attribute data DM3, the melody evaluation value EA is decreased. Specifically, the music generation unit 34 multiplies the melody evaluation value EA by a predetermined numerical value b5 (EA = EA × b5). The numerical value b5 is a positive number less than 1. Therefore, the melody data DM of the material melody suitable for the position of the unit section U in the music tends to be easily selected in the melody selection process S15.

In the sixth adjustment process, the melody evaluation value is determined according to the range difference between the melody of the melody data QU generated in the music adjustment process S16 for the unit section U immediately before the target unit section U and the material melody specified by the target melody data DM. This is a process for adjusting EA. Specifically, the music generation unit 34 adjusts the melody evaluation value EA to a smaller numerical value as the melody range difference between the previous melody data QU and the target melody data DM increases. For example, for each of the melody of the immediately preceding melody data QU and the material melody of the target melody data DM, the average value and the variance value of the pitch of each note are calculated, and the difference ΔE1 between the average values between the melody and the difference between the variance values Calculate ΔE2. The adjusted melodic evaluation value EA is calculated by multiplying the current melody evaluation value EA by the difference ΔE1 power of the predetermined numerical value b61 and the difference ΔE2 power of the predetermined numerical value b62 (EA = EA × b61 ^ΔE1). × b62 ^ΔE2 ). Therefore, there is a tendency that the melody data DM of the material melody whose sound range is similar to the melody specified by the melody data QU of the immediately preceding unit section U is easily selected in the melody selection process S15.

  The seventh adjustment process is performed when the target unit section U corresponds to the first or last unit section U of the music and the attribute data DM4 of the target melody data DM is set to the valid state (that is, the target melody data DM). The melody evaluation value EA is adjusted to a large value. Specifically, the melody evaluation value EA is updated to a value obtained by multiplying a predetermined numerical value b7 by the current melody evaluation value EA (EA = EA × b7). The numerical value b7 is a positive number exceeding 1. Therefore, the melody data DM in which the attribute data DM4 is set to the valid state is easily selected for the unit section U at the beginning or end of the music. That is, there is a tendency that music with an impressive melody added to the beginning or end tends to be generated.

  The numerical value b5 applied to the fifth adjustment processing, the numerical values b61 and b62 applied to the sixth adjustment processing, and the numerical value b7 applied to the seventh adjustment processing are the same as the numerical values b1 to b4 in the first embodiment. The variable data DX is a variable designated, and is variably set according to an instruction from the user. The melody evaluation value EA after execution of the first adjustment process to the seventh adjustment process described above is determined as the melody evaluation value EB. Note that the order of the first adjustment process to the seventh adjustment process is arbitrary. For example, the fifth adjustment process can be executed immediately before the first adjustment process SC11, and the sixth adjustment process and the seventh adjustment process can be executed immediately after the fourth adjustment process SC14.

  In the fifth embodiment, the same effect as in the first embodiment is realized. In the fifth embodiment, since the fifth adjustment process to the seventh adjustment process are executed in addition to the first adjustment process SC11 to the fourth adjustment process SC14, the character string evaluation process S14 is changed from the first adjustment process SC11 to the first adjustment process SC11. There is an advantage that a musically natural musical piece can be generated as compared with a configuration in which only each process up to the four adjustment process SC14 is executed.

<Sixth Embodiment>
In each of the above-described embodiments, a song corresponding to the designated character string DL in which each character of the lyrics created in advance is designated is generated. The designated character string DL of the sixth embodiment is composed of an array of specific characters (symbol “*” in the illustration of FIG. 12) 86 that means an arbitrary (unspecified) character as illustrated in FIG. 12. The That is, the designated character string DL designates only the number of characters in the lyrics. The user can insert a delimiter 88 at an arbitrary position in the time series of the specific character 86.

  FIG. 13 is a block diagram of a music generation device 100B of the sixth embodiment. The music generation device 100B of the sixth embodiment is preferably used to generate music that matches the designated character string DL when the lyrics are not finalized. That is, after the music is generated by the music generation device 100B, a specific song character is added to each specific character 86 of the designated character string DL to complete the song.

  As illustrated in FIG. 13, the music generation device 100B has a configuration in which the voice synthesis unit 44 and the mixing unit 46 are omitted from the music generation device 100A of the first embodiment. The acoustic signal V (musical sound signal V1) generated by the musical sound generating unit 42 according to the musical composition data Q generated by the musical composition generating unit 34 is supplied to the sound emitting device 14 and reproduced as sound waves. In the character assignment process S11 of FIG. 6, the music generation unit 34 divides the designated character string DL into a plurality of character groups G with the delimiter symbol 88 of the designated character string DL as a boundary (SA12). In the sixth embodiment, the same effect as in the first embodiment is realized.

<Modification>
Each of the above forms can be variously modified. Specific modifications are exemplified below. Two or more aspects arbitrarily selected from the following examples can be appropriately combined.

(1) In each embodiment described above, the melody evaluation value EB of each melody data DM is calculated by executing the melody evaluation process S13 and the character string evaluation process S14. However, the melody evaluation process S13 may be omitted. For example, the melodic evaluation value EA set to a predetermined value is adjusted in the character string evaluation process S14, or the melodic evaluation value is calculated from the numerical values calculated in the first adjustment process to the seventh adjustment process of the character string evaluation process S14. A configuration for calculating EB can also be employed.

(2) For example, when it is not necessary to strictly correspond each character of the designated character string DL with each musical note of the music, the melody adjustment processing S16 may be omitted. That is, the time series of the melody data DM selected for each unit section U in the melody selection process S15 can be determined as the final music data Q.

(3) The time series of notes specified by each note string data DM1 and the tonality assumed in the chord progression data DC are made common (for example, C major), and the melody specified by the music data Q generated by the music generator 34 is specified. It is also possible to change the tonality according to the instruction from the user and the ton that is musically appropriate.

(4) In each form mentioned above, although the acoustic signal V according to the music data Q was produced | generated, the production | generation of the acoustic signal V may be abbreviate | omitted. For example, the music data Q generated by the music generation unit 34 is stored in a recording medium (for example, the storage device 24 or another recording medium), or the music data Q generated by the music generation unit 34 is transmitted from the communication network to another terminal device. The structure which transmits to can also be employ | adopted.

(5) In each of the above-described forms, the chord progression data DC is generated in response to an instruction from the user. For example, a plurality of chord progression data DC corresponding to different genres is stored in the storage device in advance, It is also possible to select any one of the chord progression data DC, for example, according to an instruction from the user and apply it to the music generation process.

(6) In each of the above embodiments, all the melody data DM stored in the storage device 24 are selected as candidates for the melody selection process S15, but one of the plurality of melody data DM stored in the storage device 24 is used. It is also possible to select only part of the melody data DM as a selection candidate. For example, the melody data DM is divided into a plurality of sets (hereinafter referred to as “melody data group”) according to the number of notes and the range of the material melody, and the music for one melody data group selected from the plurality of melody data groups. A configuration for executing the generation processing (S11 to S17) may be employed. For example, a method for selecting one melody data group is arbitrary, but for example, the following method is preferably employed.

  The music generation unit 34 calculates the melodic evaluation value EB of each unit section U for each of the plurality of melody data DM stored in the storage device 24 in the same manner as in the above-described embodiments, and K unit sections The average value EB_ave of the melody evaluation value EB is calculated for U. Then, the music generation unit 34 calculates, for each of the plurality of melody data groups, the average value μ and the standard deviation σ of the plurality of melody data DM belonging to the melody data group, and out of the plurality of melody data groups. One melody data group in which the average value μ is the maximum and the standard deviation σ is the minimum (the variation of the average value EB_ave is small) is determined as a selection candidate in the melody selection process S15. According to the above configuration, since the melody data DM having a high average value EB_ave of the melody evaluation values EB and little variation is selected as a selection candidate, the effect that a natural musical piece can be generated is particularly remarkable. In the above example, the melody evaluation value EB is illustrated. However, the melody evaluation value EB can be replaced with the melody evaluation value EA.

100A, 100B: Music generation device, 12: Input device, 14: Sound emission device, 22: Arithmetic processing device, 24: Storage device, 32: Information acquisition unit, 34: Music generation unit, 42 …… Musical sound generation unit, 44 …… speech synthesis unit, 46 …… mixing unit.

Claims (5)

A device for generating a melody with a melody for each of a plurality of unit sections,
Storage means for storing a plurality of melodic data specifying different melodies;
A character string acquisition means for acquiring a designated character string composed of a plurality of characters;
A character allocating means for distributing a plurality of character groups into which the designated character string is divided to each unit section;
Character string evaluation means for calculating, for each melody data, a melody evaluation value corresponding to the number of characters of each character group corresponding to the unit section and the number of melody notes specified by the melody data for each of the plurality of unit sections. ,
A music generation device comprising: a melody selection unit that selects, for each of the plurality of unit sections, any one of the plurality of melody data according to a melody evaluation value of each melody data in the unit section. Each of the plurality of melody data includes section data specifying boundary points that divide the melody specified by the melody data into a plurality of melody sections;
The character string evaluation means associates each character group corresponding to the unit section with each melody section of the melody data for each of the plurality of unit sections, and calculates a melody evaluation value according to the number of characters for each melody section. The music generation device according to claim 1, wherein calculation is performed for each melody data. The character string evaluation means, for each of the plurality of unit sections, the larger the difference between the number of character groups corresponding to the unit section and the number of melody sections of the melody specified by the melody data, the more the melody evaluation value. The music generation device according to claim 2, wherein a melody evaluation value of each melody data is calculated so as to be a small numerical value. The character string evaluation means, for each of the plurality of unit sections, the melody evaluation value is a small numerical value when the number of characters corresponding to the unit section exceeds the number of melody notes of the melody data. The music generation device according to any one of claims 1 to 3, wherein a melody evaluation value of melody data is calculated. A melody adjustment unit that adjusts the number of melody notes specified by the melody data selected by the melody selection unit for each unit section, and associates each note in the melody after adjustment with each character in the unit section. The music generation device according to any one of claims 1 to 4.

Images