JP3364940B2 - Automatic composer - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a music apparatus, and more particularly to an automatic composer that automatically composes a melody.

[0002]

2. Description of the Related Art Conventionally, some automatic composers that automatically generate a melody are known (USP 4,399,73).
1, USP 4,664,010, WO NO.86 / 05619, JP-A-62-187
No. 876). The automatic composer of USP 4,399,731 uses a trial-and-error method in which random numbers are used to generate a melody pitch sequence as a melody generation principle. Therefore, although the melody generation space is infinite, the efficiency for obtaining a desirable melody is too low because the composition knowledge is lacking. USP 4,664,010 and WO N
The composer of O.86 / 05619 is a melody generator of the type that transforms a given melody, and a mathematical method (conversion of pitch sequence, two-dimensional space of pitch / length sequence) is used for conversion. (A linear transformation for) is used. Therefore, the conversion space is limited, and it has only fixed and mathematical (not musical) composition ability. The composer disclosed in Japanese Patent Laid-Open No. 62-187876 is a melody generator that uses a Markov model to generate pitch sequences. The melody created from the pitch transition table that represents the Markov process of the pitch sequence is bound by the musical style inherent in the transition table. Therefore, although this melody generating device can generate a musical melody with relatively good efficiency, it can generate only a limited style of melody, and the composition space that can be provided is narrow. What is common to the above prior arts is that (1) the composer lacks melody analysis and evaluation capabilities, (2) melody analysis and evaluation capabilities are not used for melody generation, (3) Therefore, the ability to generate musical melody is low,
Is.

In view of such a point, the inventor of the present invention discloses
63-250696, 63-286883, melody analysis,
We propose an automatic composer that uses the evaluation ability. This automatic composer regards the melody as musical knowledge of the melody, and considers the melody as a mixed sequence of harmony and non-harmonic sounds, and analyzes the relationship between the non-harmonic sounds and the harmonic sounds to classify the non-harmonic sounds. Has harmony classification knowledge. This non-harmonic classification knowledge is used to analyze the melody (motif) input by the user. Furthermore, the non-harmonic sound classification knowledge is also used for melody generation. This composer produces a melody in two stages. In the first stage, a harmony tone sequence (arpeggio) is generated (a harmony tone sequence is obtained by adapting an arpeggio characteristic pattern generated by developing a motif based on a motif to a music progression (chord in progress)). Non-harmonic sounds are generated in the second stage, and they are inserted between the harmony sound strings (The non-harmonic sounds are developed based on motifs and the generated non-harmonic sound characteristic patterns are adapted to the music progress (scale in progress). It is obtained by passing through the non-harmonic classification knowledge. This composer can generate musical melody that reflects the characteristics of the motif, but has the following drawbacks. (A) A considerable amount of data processing is required as a pre-processing for melody generation, in order to analyze motifs and create ideas of melody generated from motifs (feature patterns of arpeggios and non-harmonic tones). (B) The method of generating a melody in two steps is different from the normal human melody creation process. Therefore, it is impossible to realize faithful artificial intelligence in composition. (C) Real-time melody generation is difficult because of preprocessing and two-step melody generation. In fact, this composer composes in a unit of one bar in non-real time, and has a structure in which after the composition is completed, the composition is played through a sound source. According to this composition method, even if a melody idea (characteristic pattern of arpeggio and non-harmonic sound) is given in advance, it is difficult to realize the idea (melody generation) in real time. This is because the idea concrete processing (two-stage melody generation processing) is intensively performed at a specific point such as bar time, and when the composer is combined with the performance system, This is because a time delay in performance (for example, delay in sounding a melody sound) occurs when (computer) processing is concentrated. Therefore, (A) musical melody can be generated in one stage,
(B) An automatic composer capable of real-time melody generation and performance is desired.

[0004]

SUMMARY OF THE INVENTION Therefore, an object of the present invention is to provide an automatic composer capable of generating a musical melody in one stage with the support of music analysis and evaluation capabilities. Another object of the present invention is to provide an automatic composer capable of generating and playing a melody on a real time basis.

[0005]

According to the present invention, a music progression giving means for giving a music progression and a rule base of a melody pattern in which a tone sequence of a melody is patterned by a tone type sequence and a tone sequence are stored. and the melody pattern rule-based means for the eyes that can be melody input by the user
It is input by the body input means and this melody input means
Add the sound type of each sound of the melody by the music progression adding means
Based on the progress of music being recorded,
By determining the pitch of the melody sound that
A melody that creates a melody pattern expressed as a sequence
Pattern creating means and this melody pattern creating means
The melody pattern created in
Melody pa to add to rodi pattern rule base means
An automatic tune comprising: a turn rule base expanding means , and a melody generating means adapted to generate a melody that matches the music progression from the music progression giving means and that satisfies the melody pattern rule in the melody pattern rule base means. There is provided an automatic composer having a machine. According to this configuration, it is possible to generate the melody without dividing into the generation of the harmony tone sequence and the addition of the non-harmonic tone sequence. Further, since the generated melody is suitable for the music progression which is the musical background of the melody and also for the rule (melody pattern rule) placed in the melody pattern rule base means, this configuration is a musically preferable melody. Can be provided relatively efficiently. That is, the characteristics of the generated melody evaluated by a musical progression (context) of the melody (for example, a tonal harmony background such as an ongoing chord, key, or scale (note sequence and pitch). columns) that embodies (applied) is the melody in accordance with the musical context of the melody to be composed of. in other words the melody pattern rule satisfies the melody pattern rule. Moreover, this arrangement is described in the prior art JP-A-63
-250696, it is possible to generate a melody with a much smaller amount of data processing than the automatic composer disclosed in Japanese Patent Laid-Open No. 63-286883. Therefore, it facilitates the real-time basis of melody generation and performance. Furthermore, this structure
In addition, the melody pattern rule is added by the user.
You can add a melody that suits your taste.
The ability of the I generation can be high Mel, to the user of the composer
Enable active participation.

According to another aspect of the present invention, a melody tone
The automatic music composer that automatically generates
And a means for adding music progression,
Rule base of melody pattern patterned with
Melody pattern rule-based means to store
Stored in the melody pattern rule base means of
Without referring to the melody pattern rule base, the current
Melody sound candidate generator for generating existing melody sound candidates
And the current song generated by this melody sound candidate generation means.
The sound type of the rody sound candidate is assigned by the music progression assigning means.
Sound type determination means for determining based on the progress of music,
Current melody tone generated by melody tone candidate generation means
The pitch between the complement and the previously generated melody sound is determined.
Pitch determination means to be determined and the sound type determined by the sound type determination means
And a past tone series including the pitch determined by the pitch determination means
And the pitch sequence pattern is the melody pattern rule base.
Rule base to determine whether or not it is stored in
Stored by the search means and this rule-based search means
If it is determined that the current melody sound candidate is
Followed by this melody sound as it was decided as a Roddy sound
Next melody sound candidate generation means for generating melody sound candidate
And the rule-based search means
When judged, the new melody sound alternative
Instruct the melody sound candidate generation means to generate melody sound candidates
Automatic composer comprising:
Will be provided. With this configuration, the melody generation is
Time series without dividing into voice sequence generation and non-harmonic sequence addition
Can be performed in sequence, and can generate melody on a real-time basis.
And performance is easy. Is also generated
Melody is a musical background of the melody
And rules placed in melody pattern rule base means
Since it is compatible with
Can be generated.

[0007]

[0008]

Embodiments of the present invention will be described below with reference to the drawings. Features Automatic composer (automatic melody generator) according to the present embodiment
The main features of are the composition function of composition, the function of generating pitch sequence of melody, and the expansion function of melody pattern rule.
The features of each are described below. Composition Conditioning Device (FIG. 1A) FIG. 1A shows a functional block diagram of the composition condition device 100 incorporated in the embodiment. The music composition device 100 includes a music structure database 110 and a chord progression database (CP
Generate a selected chord progression (CP) 160 based on DB). Furthermore, the music composition condition device 100 uses the music structure data 11
0, song style input 130 and rhythm pattern database 1
A selection rhythm pattern 170 is generated based on 50.
The selected chord progression 160 column represents the chord progression of the song being composed. The row of the selected rhythm pattern 170 represents the rhythm (sound length row) of the melody to be composed. The selected chord progression 160 is obtained by taking a chord progression suitable for the selected music structure 120 from the chord progression database 140. The selected rhythm pattern 170 is selected from the rhythm pattern database 150.
It is obtained by extracting the rhythm pattern that matches 30. In FIG. 1A, the music structure database 110
Consists of an upper structure database 111 and a lower structure database 112. The superstructure database 111 stores a database of superstructures of music (for example, passages). The substructure database 112 stores a database of a substructure of music (for example, a passage internal structure). For example, each floor structure entry of the lower structure database 112 has information indicating an 8-bar structure when the upper structure is regarded as a 8-bar passage. Selection module 116 takes the upper structure data from the upper structure database 111, and stores the selected music structure 120 it as the selected upper structure 121. Aptitude module 145 removed from the chord progression database 140 chord progressions that match the selected upper structure 121, and generates it as the selected chord progression 160. The aptitude module 117 extracts a lower structure column (for example, an 8-bar structure) having an attribute that matches the selected upper structure 121 from the lower structure database 112 , and selects it from the selected lower structure 12.
2 is set in the selected music structure 120. The selection substructure 122 is provided to the fitness module 155 for attribute testing. The musical style input 130 includes a designated rhythm 131 and a designated beat (or tempo) 132. Designated rhythm 13
Both 1 and the designated beat (or tempo) 132 are provided to the fitness module 155 for attribute testing. The suitability module 155 matches the selected substructure 122 from the rhythm pattern database 150, and specifies the rhythm 13
1 and find a rhythm pattern having a matching attribute specified beat (or tempo) 132, and generates it as the selected rhythm pattern 170. The selected rhythm pattern 170 generated in this way defines the tone length sequence of the automatic melody (pronunciation and muffling timing of each melody tone). In this way, the music composition condition device 100 uses various databases as knowledge resources and music information (selection chord progression, selection rhythm pattern) that is appropriately conditionalized based on the music style input from the user as a part of composition. To generate.
Since the tone length sequence of the melody is obtained from the selected rhythm pattern 170, the melody is completed by generating the tone sequence of the melody. The music structure data 11 of FIG. 1A
In the case of 0, the hierarchical levels of the structure are upper and lower, but the number of hierarchical levels may be 1 or 3 or more. Therefore, the number of hierarchical levels of the selected music structure may be one or more and may be any number. Although the musical style input is given by the designated rhythm 131 and the designated beat (tempo) 132 in FIG. 1A , other musical style parameters may be used.

Pitch Sequence Generating Device (FIG. 1B) FIG. 1B is a functional block diagram of a melody pitch sequence generating device 200 incorporated in the automatic composer. The purpose of the pitch string generation device 200 is to generate a pitch string of a melody. The pitch string generation device 200 can be combined with (but not necessarily limited to) the composition / conditional device 100 as shown in FIG. 1A. In this case , the pitch string generation device 200 generates the determined current pitch 224 each time the selected rhythm pattern 170 indicates the sounding timing of the melody sound. The row of the determined current pitch 224 defines the pitch row of the melody. When generating the determined current pitch 224 in FIG. 1B, the pitch train generation device 200 adds the previous pitch 202 and the random number from the random number generation module 204 in the adder 206 to generate the first current pitch candidate Markov. A tone type / pitch sequence generation module 20 to which the current pitch candidate is input
8 to generate the column of note type and pitch of the pitch sequence plus the current pitch candidate in the generated pitch sequence. The tone type / tone sequence generation module 208 uses the information of the current tonality 210 (key, scale) and the current chord 212 to classify the tone type and tone of the current pitch candidate. The output of the tone type / tone string generation module 208 represents the pattern of the melody tone string up to the current pitch candidate, and the information thereof is given to the matching module 214. Matching module 214 gives a melody
(Sound sequence) pattern Melody pattern rule (data
Database) 216. As a result of the search, when the melody pattern rule base 216 does not include an entry corresponding to the melody pattern defined by the output of the module 208, the search result determination module 218 is NO, and the next candidate generation module 220 generates the next current pitch candidate. To be done. The next candidate generation module 220 generates the next current pitch candidate using the information of the previous pitch + random number shown in 222. The operations of the tone type / tone string generation module 208 and the matching module 214 are repeated again, and the validity of this new current pitch candidate is evaluated. If a melody pattern corresponding to the output of the tone type / tone string generation module 208 is entered in the melody pattern rule base 216 , the search result determination module 218
The answer is YES, and the current pitch candidate at that time is generated as the determined current pitch 224. When the next melody tone is generated, the determined current pitch 224 becomes the previous pitch 202, and the above-described operation is repeated to determine a new current pitch.
In this way, the pitch string generation device 200 can sequentially generate the pitches of the melody in real time by using the melody pattern rule base 216. Therefore, the output of the pitch train generation device 200 (determined current pitch 2
Only contains the real-time performance-based 228 (electronic sound source 24), rhythm
Pattern is given ) , composition and performance can be executed simultaneously in real time. That is, the pitch train generation device 200 has a real-time composition capability.

Melody Pattern Rule Expansion Device (FIG. 1C) FIG. 1C shows a functional block diagram of the melody pattern rule expansion device 300. Melody pattern rule expansion device 3
00 is a melody pattern rule base that is referred to by a pitch sequence generation device (automatic melody generation device) as shown in FIG. 1B.
The function to extend 216 is realized. Of the melody pattern rule (data) base 330 (216) , 330F is a fixed melody pattern rule base fixedly incorporated in the automatic composer, and 330E is an extension of the melody pattern rule base. According to the melody pattern rule expansion device 300, the melody pattern rule of the expansion section 330E can be created based on the user input melody 302. The part that creates this rule is the melody pattern (tone type and pitch sequence)
The generation block 310. User input melody 30
2 is input to the pitch classification module 312 and the tone type classification module 316 of the melody pattern generation block 310. The pitch classification module 312 evaluates pitches formed between adjacent sounds of the user input melody 302 to form pitch strings. For example, as shown in 314, pitch classification module sound column to classify the pitch as such homophones (Motion no) or ascending order progress or descending order progress or ascending leap progress whether descending leap progress. 316 classifies the sound types of the sounds of the user input melody 302 to form a sound type sequence. The tone type classification module 316 uses the information of the key scale code 318 and the standard pitch class set (PCS) memory 32 of each tone type to classify the tone type of the melody tone.
Use 0 information. For example, as shown in 322, the note type classification module 316 classifies the note type as such note type code or tone or scale notes or A Vu Eiraburunoto or tension notes or Avoidonoto melody sound. The pitch sequence output of the pitch classification module 312 and the pitch string output of the pitch classification module 316 represent the pattern of the user input melody 302, that is, the rule of the melody pattern derived from the user input melody. This rule is an extension 330E of the melody pattern rule base.
Memorized in. When the melody is generated, the automatic melody generation module 340 (for example, the pitch string generation device 20 of FIG. 1B) is used.
0) uses the entire melody pattern rule base 330 including the extension section 330E to generate a melody. By providing such a melody pattern rule expansion device 300, it is possible to provide an automatic composer having a high ability to generate a melody desired by a user.

Hardware Configuration (FIG. 2) FIG . 2 shows a typical hardware configuration that realizes the functions described in FIGS. 1A, 1B and 1C. CPU2 is R
The entire system is controlled according to the program stored in the OM4. The ROM 4 stores various fixed data (such as the database described above) in addition to the program . The RAM 6 stores various variables and temporary data, and is used as a working memory of the CPU 2. Keyboard 8 can be configured in a conventional electronic keyboard musical instrument of the keyboard, it is used for the user of the melody input. Input device the start of the musical style input device and automatic melody generation to 10, stop input device or the like including Murrell. The states of the keyboard 8 and the input device 10 are appropriately read by the CPU 2 and the corresponding processing is performed. The sound source 12 generates a tone signal under the control of the CPU 2. Sound system 14 is sound source 1
Receiving the tone signal from the 2, it reproduces the sound. The display device 16 displays system functions and messages under the control of the CPU 2. The display device 16 includes an LED and a liquid crystal display. The clock oscillator 18 generates a pulse every time the music resolution time elapses, and the CPU 2
To execute an interrupt program (interrupt routine described later).

Automatic Melody Generation Function (FIG. 3) FIG. 3 shows a block diagram of the automatic melody generation function incorporated in the configuration of FIG. The automatic melody generation function has an input means 20 for inputting information necessary for melody generation, a means 30 for generating music data, an analysis and evaluation means 40 for analyzing and evaluating a melody, a clock oscillator 60, a rhythm counter 7.
0, a melody memory 80, and a sound source 50. Input means 2
Rhythm selector 2 for inputting designated rhythm type RHY to 0
1, a keyboard device 22 for inputting a key code KC, and a tempo setting means 23 for inputting a designated tempo TMP. In addition,
The input means 20 also includes means for instructing the start / stop of automatic melody generation in addition to this, but the illustration is omitted. Generation means 3
At 0, the structure generation means 32 uses the music structure database 3
Extract music structure data from 1. The extracted music structure data is composed of upper structure data and lower structure data matching it. Code generating means 34 extracts the chord progression that fits the music structure and musical style from the chord progression database 33 (designated rhythm species RHY), and outputs the code CHO at the current time T from the rhythm counter 70. Rhythm pattern generation means 36 uses the rhythm pattern database 35
Music structure and style (specified rhythm type RHY and specified tempo TM
The rhythm pattern PAT that matches P) is taken out. Specifically, the designated tempo TMP is subjected to beat type BEAT conversion by the tempo / beat correspondence map 48, and this is given to the rhythm pattern generation means 36 as one of the conditions for the rhythm pattern generation in the rhythm pattern generation means 36. Thus, the chord generating means 34, the structure generating means 32, and the rhythm pattern generating means 36 have the function of the suitability module as described in FIG. 1A. The random number generating means 38 gives the pitch data generating means 39 the random number data from the random number memory means 37 as RAN. The pitch data generation means 39 uses the random number data RAN and the previous pitch to select a current pitch (pitch) candidate.
Generate PIT . Note type classification means 42 in the assay unit 40 receives the (output of the output or the keyboard device 22 of the tone pitch data generating means 39) (local procurement of) key and (now) code CHO and pitch data, the sound The seed NT is classified by referring to the standard pitch class set memory 41. The pitch classifying unit 43 classifies the pitch formed between the previous pitch and the current pitch (candidate) as M T. The classification results of the pitch type classifying unit 42 and the pitch classifying unit 43 are stored in the pitch type and pitch string memory 47. The fixed melody pattern database memory 45 is stored in the ROM 4 of FIG. The extended melody pattern database memory 46 stores melody pattern data (rules).
It is stored in the RAM 6 for expanding the base. Evaluation means 4
4 searches the fixed melody pattern database memory 45 and the extended melody pattern database memory 46 for the tone type and tone sequence and tone sequence (melody pattern) from the tone sequence memory 47, and if there is a corresponding rule as a result of the search , OK is determined . The evaluation result JDG is output. This determines the current pitch PIT (or KC) . Rhythm counter 70 outputs the present time data T by counting <br/> is that the clock pulses supplied from the clock generator 60 for each music resolution time. Melody memory 80 has PIT (pitch) and T
The data of the melody tone sequence that is the sequence of (timing) is stored. Source 50 each of tone generation timing of the melody tone, receives the pitch data PIT, generates a corresponding musical tone signals.

Expression Format of Various Music Elements (FIGS. 4 to 6) In the following description of the embodiments (especially, the description of the flow), various music elements mainly use a character expression (symbol expression), and if necessary, a numerical expression. Is used. Correspondence between the character representation and the numerical representation (machine level representation) of various music elements will be described below with reference to FIGS. The character representation of the code type is CHO (TYPE). Code type MAJ is number 0, MIN is number 1, 7TH is number 2, M
IN7 is numerically expressed as 3 and MAJ7 is numerically expressed as 4. The data length of the code type in the machine is 6 bits, for example, and it is possible to represent a maximum of 64 code types. The character representations of the chord root and the key are CHO (ROOT) and KEY, respectively. A chord root or key is specified by its pitch class (note name) . The pitch class C is numerically expressed as 0, C # is 1, D is 2 , and so on, B is numerically expressed as 11. Therefore, the effective data tone of the chord root and the key is 4 bits. The upper music structure in the music structure (upper / lower) is S
It is expressed in TR1D and the lower music structure is expressed in STR2D. Higher music structure of AA is 1H, BB is 2
H and CC are numerically expressed by 4H. Lower music structure aa is numerically expressed by 1H, bb is 2H, cc is 4H. In the apparatus of the embodiment, the effective length of the structure data is represented by 4 bits (nibbles). Rhythm type is RHY
Rhythm type ROCK (rock) expressed by letters is 1H,
DISC (Disco) is 2H, 16BE (16 beats)
Is 4H, SWIN (swing) is 8H, and WALT (waltz) is 10H. The pitch data and the key code are represented by PIT and KC, respectively. The pitch data and the key code are specified by the pitch class and octave. In the embodiment, the numerical value 0 represents the pitch of C2 (the pitch class is C and the octave is the second octave), and the pitch is expressed by the numerical value which is incremented by 1 each time the pitch increases by one semitone. Note type numerical note type code tones character representation (Chot) in NT 0, A Vu Eiraburunoto (AvaI) numeric 1, scale notes (SCA
L) is a numerical value, tension note (TENS) is a numerical value 3, a void note (AVOI) is a numerical value 4, end mark (END) is a f H, pitch symbolic expression is MT, same tone ( (SAME) is the number 0,
Ascending sequential progress (+ STEP) is number 1, downward ascending sequential progress (-STEP) is number 2, ascending leap progress (+ JUNP)
Is represented by the numerical value 3, the downward jumping progress (-JUNP) is represented by the numerical value 4, and the delimiter mark (SEP) is represented by f H. The evaluation result is expressed in characters by JDG (flag JDG), the evaluation result of NG is expressed by the numerical value 0, and the evaluation result of GOOD is expressed by the numerical value 1. USER_MEL expresses NO by the numerical value 0 and YES by the numerical value 1. The elements of the pitch class set (pitch class) are expressed as follows. Pitch class C is 1
H, C # is 2H, D is 4H , and so on, B is 800
Expressed as H. That is, the pitch class is represented by the bit value "1" at the bit position in 12 bits (for example, the least significant or 0th bit position is C). A pitch class set, which is a set of pitch classes, is expressed by a bitwise logical sum between such 12-bit data expressions. For example, a pitch class set consisting of C and D is expressed by a logical sum of 1H and 4H, that is, 5H. Each character representation of the pitch class set pitch class set of code tones are PCS (CT), in place of CT T N, SN, A
N, the use of the AV, the tension notes respectively, scale note, A Vu Eiraburunoto, characters representing a pitch class set Avoidonoto. The character representation of the rhythm pattern is PAT. The element of the rhythm pattern is 1
It is expressed by 6-bit data, and each bit position is 1 bar.
Each timing when dividing into 6 is shown. The rhythm pattern consists of a key-on pattern (sound generation timing pattern) and a key-off pattern (silence timing pattern). The sound generation timing or the mute timing on the bar line is numerically expressed by 1H. The sounding / silence timing at 1/16 bar after the bar line is expressed by 2H.
The same applies hereinafter. A key-on pattern or a key-off pattern, which is a set of elements of a rhythm pattern, is expressed by a bitwise logical sum of 16-bit data elements.
For example, a key-on pattern in which the first beat and the second beat are the sounding timings in the case where one measure has four beats is expressed by a logical sum 11H of 1H and 10H. The beat type is 4 BEA
T is 1H, 8BEAT is 2H, 16BEAT is 4H.
Each is expressed numerically.

Storage Format of Various Data (FIGS. 7 to 14) Next, examples of various data will be described with reference to FIGS. 7 to 14. The data length (1 word) is 16 bits. The name of the data memory and the start address (start symbol address) of the data memory are called with the same name. For example, the chord progression data memory is called CPM, and the head address of the chord progression data memory is also called CPM. FIG. 7 shows a standard pitch class set database memory. Standard pitch class set database memory is code tone memory CT, tension notebook memory T N, scale notes menu
It consists of Mori SN. Chord tone memory CT receives the chord type is a table that returns a standard pitch class set of the code type. For example, the data 91H at the symbol address MAJ of the code tone memory CT indicates that the standard pitch class set of the chord type MAJ is C, E, G. Here, the standard pitch class set is a pitch class set when the chord root is C. When conversion words, the standard pitch class set database memory is a representation of a pitch class set for each note type in a relative pitch (pitch structure). The tension note memory T N is a table for changing the standard pitch class set of tension notes for chord types. For example, for the chord type MAJ, the tension note memory PN returns data a 44H indicating that D 1 , F #, A and B are standard pitch class sets. Scale note memory SN is rhythm type (or scale type)
It is a table that returns the standard pitch class set database memory of scale notes. For example, scale note memory S for rhythm type rock (ROCK)
N is the standard pitch class set database memory of the scale note in which the data ab 5H is replaced by C, D , E,
Notify that it is F, G, A, B. FIG. 8 shows the fixed melody pattern data (rule) base memory MPM1.
And the tone type and pitch data memory NTM and MTM . The fixed melody pattern database memory MPM1 is shown in FIG.
It is stored in ROM4. The memory MPM1 realizes a melody pattern rule base, and each rule is represented by 2 words (32 bits). Each rule consists of a note type string and a pitch string. For example the first rule memory MPM1 upward from the chord tone (Chot) (ascending) sequence proceeds (+ STEP) by available note (A
VAI), and an upward sequence (+ STEP) from an available note to a chord tone (CHOT). In the 2-word rule data, each tone type and pitch is represented by 4 bits (1 nibble). Therefore, in the case of this storage format, the number of sounds of the melody pattern expressed by the rule is four at maximum. The SEP nibble of the pitch sequence word that is the second word of the rule is the end mark (separation mark between rule entries) of the rule. A code END indicating the end of the fixed melody pattern rule base is provided at the final address of the memory MPM1.
Is written. A memory MPM2 having a storage format similar to that of the MPM1 is provided in the RAM 6 for expanding the melody pattern rule base.

The tone sequence memory N T M and the tone sequence memory M T M
Represents an input melody pattern (a melody pattern derived from a user input melody from the keyboard 8 or a melody pattern derived from an automatically generated melody). N T M and M T M are both formed in RAM 6 1
In the word (16-bit) tone string memory N T M, each tone type is represented by 1 nibble (4 bits), and similarly, in the 1-word (16-bit) tone string memory M T M, each tone data item is It is represented by 1 nibble (4 bits). Therefore, N T M and M T M represent an input melody pattern for up to four tones. As will be described later, for generation of a new melody sound (automatic generation or input of a new melody sound from the keyboard 8), N T M and M T M are operated as right shift registers in units of nibbles. During production start, or the contents of the NTM and MTM at the timing of a bar line at the right end of the nibble is initialized so that END nibbles. Therefore NTM and MTM shown in FIG. 8 after the initialization, and shows a state that contains the pattern up to three sounds. In the expansion processing of the melody pattern rule base, the contents of NTM and MTM are added to the expansion melody pattern rule base in the RAM 6 as additional rules. In the automatic melody generation process, the contents of NTM and MTM are used to search the melody pattern rule base for the presence or absence of the corresponding rule entry. Tempo-bi in Figure 9-1 door conversion map T
BTM and random number memory RANM are shown. TBTM and RAN
Both M are stored in ROM4. Tempo bi over preparative conversion map TBTM is a table that returns a beat type for the specified tempo from the input device 10 (e.g., 16 bits).
Output tempo bi over preparative conversion map TBTM is used to conditionally rhythm pattern as described in Figure 1A. The random number memory RANM stores random number data. Random number memory R
The ANM is used to obtain a pitch candidate of a new melody tone in the melody automatic generation mode. Random number memory R
An end mark (ffffH) indicating the end of random number data is stored at the final address of the ANM. FIG. 10 shows the chord progression database memory CPM. This memory CP
M is stored in the ROM 4 of FIG. The memory CPM realizes a database of chord progressions. Each chord progression entry in the database CPM consists of the following data items. The first word represents the attribute of the rhythm type. For example, the attribute data of 00111 indicates that the chord progression has attributes suitable for rhythm type rock (ROCK), disco (DISC), and 16 beats (16BE). The rhythm type attribute data item of the chord progression entry is shown in FIG. 1A.
It is used for the attribute test with the specified rhythm type as described in. In the second word of the chord progression entry, the length is stored in the upper part and the music structure attribute is stored in the lower part. For example, if the music structure attribute data item is 101 , the chord progression has an attribute that matches the superstructure AA or CC, but the superstructure B
B has an attribute that is not suitable for B. The music structure attribute data item is used in the chord progression attribute test for the selected superstructure as described in FIG. 1A. The data body of the chord progression is stored in the third and subsequent words of the chord progression entry. Each word stores the code root in the upper word (1 nibble), the code type in the middle (2 nibbles), and the code length in the lower (1 nibble). The word 0000H is stored at the end of each chord progression entry as a separator (separation mark) between the entries. At the end of the chord progression database memory CPM is stored the word ffffH representing the end mark of the database.

FIG. 11 shows a music structure database memory. The music structure database memory is located in ROM 4 of FIG. The music structure database memory includes an upper structure database memory STR1 and a lower structure database memory S.
It consists of TR2. High-level structure database memory STR
Reference numeral 1 stores a database of higher-order structure sequences representing the structure of a piece of music as a phrase structure sequence. Therefore, for example, the first nibble in which the upper structure sequence data of up to 4 movements is stored in each word is A
A, the second nibble is BB, the third nibble is AA, and the fourth nibble is END, the musical composition word is AA for the first verse, BB for the second verse, and AA for the third verse (final verse). .
A word ffffH indicating the end of the database is stored at the end of the upper structure database memory STR1. The substructure database memory STR2 has a passage internal structure (8
Store a database of bar structure). Each substructure entry of the substructure database memory STR2 consists of 3 words, the first word is assigned to the attribute and the second and third words are assigned to the structural representation of 8 bars. The attribute word (first word) has information indicating a higher-order structure to which the entry internal structure (structure sequence of 8 measures) conforms. For example, the attribute data item 1100 indicates that the internal structure of the phrase is compatible with the CC or DD phrase but not with the AA or BB phrase. The attribute data in the lower structure database memory is used in the attribute test for selecting the lower structure that matches the upper structure as described in FIG. 1A. Substructure entry 2nd and 3rd
The structure of each bar in a word is represented by one nibble (4 bits). In detail, the first nibble (most significant 4 bits) of the second word is the structure of the first measure, the second nibble is the structure of the second measure, and the like, the first nibble of the third word is the structure of the fifth measure. , 4th nibble of 3rd word (least 4 bits)
Indicates the structure of the eighth measure. The word ffffH indicating the end mark is stored at the final address of the lower structure database memory STR2.

FIG. 12 shows the rhythm pattern database memory RPM. Rhythm pattern database memory R
The PM is stored in the ROM 4 of FIG. The rhythm pattern database memory RPM realizes a database of automatically generated melody rhythm patterns (note length sequences). Each entry of the rhythm pattern database memory RPM is made up of 4 words, the first word stores the attribute of the rhythm type, the second word stores the attribute of the beat type in the upper part, and the attribute of the lower structure of the music in the lower part.
The word represents the key-on pattern and the fourth word represents the key-off pattern. If the rhythm type attribute data item in the first word of the entry is 00111, the rhythm pattern of the entry is rhythm type lock (ROC
K), disco (DISC), 16 beats (16BE)
Suitable for, but swing (SWIN) and waltz (WA
It is not suitable for LT). Moreover, indicating that structural attribute data item to be placed in the second word subordinate rhythm pattern entry if 101, but the rhythm pattern of the entry is suitable for measure the structure of aa or cc unsuitable for bar structure bb . The key-on pattern stored in the third word of the entry describes the sounding timing of each note in one bar. The key-off pattern in the 4th word describes the muffling timing of each note within one bar. The word ffffH representing the end mark is stored at the final address of the rhythm pattern database memory RPM. The attribute data in the first and second words of each rhythm pattern entry is selected from the rhythm pattern database as described in FIG. 1A. structure, specified rhythm, is used in the attribute test of when searching for a rhythm pattern that matches the specified beat (tempo). Other constants and variables used in the embodiment will be described with reference to FIGS. 13 and 14. MPM2 represents an extended melody pattern data (rule) base or its head address played by the user. The extended melody pattern database MPM2 is created in the RAM 6 of FIG. 2, and MPM2SIZE represents the memory size of the extended melody pattern database.
MEAS represents the length of one bar . CP_C is a chord progression length counter (code length accumulator) . T is a rhythm counter indicating the current time . BAR
Ri is bar counter der, the value of the bar counter BAR T /
It is defined by the integer part of MEAS . T MP represents a specified tempo. BEAT represents a designated beat type and is obtained by passing the designated tempo T MP through the tempo / beat conversion map. STR 1D represents an upper structure (eg, AA passage structure) selected from the upper structure database memory STR 1, and STR 2D represents a lower structure (bar structure, eg, a) selected from the lower structure database memory STR 2.
It represents a). P IT represents the current pitch and PREV_
PIT represents the front pitch.

STR1_P is a word (address) pointer of the upper structure database . STR2_P is a word (address) pointer of the substructure database. STR 1 D _P is higher structural element counter (passage counter in the word representing the song structure in the upper structure database memory). STR2 D _P is a counter of a substructure element (a bar counter to an 8-bar structure which can be the second and third words in the entry of the substructure database memory). The counter STR 1D _P is used to retrieve the phrase structure at the current time from the music structure word selected from the database STR1. Counter STR 2D _P whereas the selected entry from the database STR 2, used when taking out the measure structures in the current time. RPM_P is a word (address) to the rhythm pattern database memory RPM
It is a pointer. CPM_P is a word (address) pointer to the chord progression database CPM. RAN
M_P is a word (address) pointer to the random number data memory RANM. MODE represents the mode of the automatic composer. NO in normal mode that does not automatically generate melody
The mode indicated by RMAL and in which the melody is automatically generated is indicated by RMELODY . KEYON the Ri flag der key depression state, the flag KEYON is YE key depression state
S and NO indicate the key release state . PKEYON is a flag indicating the pressed key state at timing immediately before the pressed key state Y
ES and key release state are indicated by NO. USER_MEL is a flag indicating whether or not the user's performance is within the bar. This flag USER_MEL is an automatic generation mode RM
In ELODY, it is used as a flag for setting a melody pattern rule expansion mode for adding the melody pattern by the user's melody performance to the melody pattern database (the automatic generation of the melody is prohibited during that time).

Outline of operation of the embodiment The automatic composer of the embodiment composes in real time, and performs the composition in parallel with the composition work in real time. In response to the generation start instruction from the input device, the operation mode of the automatic composer shifts to the automatic melody generation mode, and in response to the generation start instruction, the automatic composer determines the structure of the music composition. This decision is initiated by retrieving the song structure word from the music structure database. Furthermore, the structure of the first passage contained in the song structure word is selected as the structure at the start of generation, and the internal structure of the passage (bar structure, substructure) suitable for the selected passage structure is retrieved from the substructure database ,
Select that first bar structure as the first bar structure of the song. Furthermore, the automatic composer is suitable for the structure of the passages (upper structure) from the chord progression database , and can be specified by the input device.
The chord progression suitable for the song is extracted, and the first chord of the chord progression is determined as the first chord of the song. Further, the automatic composer retrieves from the rhythm pattern database a rhythm pattern that matches the bar structure (subordinate structure) and that matches the specified rhythm from the input device and the specified beat. The extracted rhythm pattern has a length of one measure, and the sounding timing information and the mute timing information of the automatically generated melody sound in that one measure are key-on patterns,
It is written as a key-off pattern. Information at each music point is read from the rhythm pattern extracted as the music time progresses in accordance with the designated tempo. The automatic composer determines the pitch of the melody tone when the read information indicates the timing of the automatically generated melody tone. The automatically generated melody sound is determined as follows. Automatic composer retrieves random data from the random number data memory, it makes the pitch candidates of the previous melody tone added (or subtracted) Te melody tone to be newly generated. Next, the automatic composer creates a melody pattern (defined by the tone type sequence NTM and the tone sequence MTM) up to the pitch candidate, and regarding the melody pattern, the melody pattern rule (data) base MP M 1,
Search for MP M 2. If the search fails, another pitch candidate is generated and the search is repeated. If the search is successful, the candidate at that time becomes the pitch of the new melody tone. The pitch data of the melody tone newly generated in this way is reproduced (played) as a sound through the sound source. When the timing information of the rhythm pattern at the current time indicates the mute timing of the melody sound, the melody sound being sounded is muted via the sound source. Since the rhythm pattern fetched from the rhythm pattern database is one bar long , the composer again fetches an appropriate rhythm pattern suitable for the progress of music from the rhythm pattern database every time the music time of one bar elapses. That is, a rhythm pattern that matches the specified rhythm and specified beat and that matches the structure of the new measure is extracted from the rhythm pattern database. The composer is
The chord progression retrieved from the chord progression database is scanned according to the current time from the rhythm counter. In this embodiment, the length of chord progression is the same as the length of the higher-level structure (section). When the chord progression being selected ends (when one passage ends), the automatic composer retrieves the next passage structure (upper structure) from the song structure word, and selects the chord progression that matches the passage structure from the chord progression database. Take it out.
The melody is automatically generated as described above. In addition,
It is easy to perform automatic accompaniment based on chord progressions retrieved from the chord progression database. In the automatic composer of the present embodiment, when the user inputs a melody in the automatic melody generation mode , the melody is patterned (ruled) and added to the melody pattern rule base as an additional rule. Can be done. This is called a melody pattern rule extension function. That is, when the first key depression occurs on the keyboard within one bar, the mode of the automatic composer is the melody rule extension mode (USER
_MEL = YES). Automatic composer is to the user melody sound from the keyboard, to evaluate the sound type and pitch. The evaluated pitch type and pitch are loaded into the pitch string memory MTM and the pitch string memory NTM by the right shift method, respectively. At the timing of the next bar line, the NTM and MTM store melody pattern information which is a rule of the user's melody input during one bar. Therefore, the automatic composer writes the melody pattern information in the NTM and MTM into the extended melody pattern rule base MPM2 to realize the extension of the melody pattern rule base. At the bar line timing, the melody pattern rule-based expansion mode is automatically canceled and the melody is automatically generated to return to the automatic generation mode.

Description of Flow Hereinafter, the embodiment will be described in more detail with reference to FIGS. Main Flow (FIG. 15) FIG. 1 shows a flow of a main routine executed by the CPU of FIG.
5 shows. First, the main routine initializes the system in step 15-1. Main loop 15-2 to 15-
A key scan is executed at entry 15-2 of 16 to read the states of the keyboard 8 and the input device 10. Then, required processing is executed corresponding to each operated key . That is, when there is a generation start instruction (step 15-3), the generation start processing 15-4 (FIG. 18) is executed, and when there is a stop instruction (step 15-5), the stop processing 15-6. (FIG. 24) is executed, and when a key is pressed on the keyboard (step 15-7), a key pressing process 15-8 (FIG. 25) is executed, and when a key is released on the keyboard (step 15-9). ), The key releasing process 15-10 (FIG. 26) is executed, and when the rhythm switching is instructed from the input device 10 (step 15-11),
Rhythm switching 15-12 is executed to execute rhythm type register RH
When the rhythm type designated in Y is set and the tempo setting is input from the input device 10 (15-13), the tempo setting 15-14 is executed to set the tempo designated in the tempo register tempo, and other For the input of (step 15-15), other processing 15-16 is executed. Interrupt Routine (FIG. 16) FIG. 16 shows the flow of the interrupt routine. Interrupt routine signal representing the course of the music-resolution time from the clock oscillator 18 of FIG. 2 is activated each time that occurs. First, in 16-1, it is determined whether MODE = NORMAL , that is, whether the mode is normal. If the mode is normal , the process directly returns , and if it is not normal (in the automatic melody generation mode), the following processing is performed. That is, in step 16-2, it is checked whether or not the key is being pressed, and if the key is being pressed, the key pressing process 16-3 (FIG. 27) is executed and then the process proceeds to step 16-9. If the key is not being pressed , it is determined in step 16-4 whether USER_MEL = YES , that is , whether there is a melody input by the user at the current measure (whether the melody pattern rule base expansion mode is set), and USER_MEL = YES remains unchanged. Proceed to step 16-9, USER_MEL =
If NO is established, the process proceeds to step 16-5 and Get.
B (* (RPM_P + 1), TmodMEAS, 1) =
It is determined whether it is 1 or not , that is , whether it is the sounding time of the automatically generated melody sound . If it is the sounding time of the automatically generated melody sound, the processing 16-6 (FIG. 33) of the sounding time is executed and then the process proceeds to step 16-9 . If it is not pronunciation time, the process proceeds to step 16-7, GetB (* (RPM_P + 2),
It is determined whether TmodMEAS, 1) = 1 , that is , whether it is the time to silence the automatically generated melody sound . If it is the mute time of the automatically generated melody sound , mute time processing 16-8
After executing (FIG. 35), the process proceeds to step 16-9, and if it is not the muffling time, the process immediately proceeds to step 16-9. In step 16-9, Inc CP_C, T PKEYON
= Running KEYON, updated code counter of progress, the rhythm counter, Utsushitoru the contents of the current key state register just before the key state register. Next, step 16-1
0 to determine whether the end of the chord progression, the running <br/> lever steps 16-11 have reached the end, the upper structure generation processing (FIG. 19) and the chord progression generating process (FIG. 2 2) To do .
And whether or not the small section line position in the step 16-12 Tm
It is determined from odMEAS = 0, and if it is a bar line time, steps 16-13 are executed, and the lower structure generation processing (see FIG.
0) , rhythm pattern generation processing (FIG. 23) and melody pattern DB expansion processing (FIG. 36).

Initialization (FIG. 17) FIG. 17 shows details of the initialization 15-1 of the main routine. 17-1: STR1_P = STR1, STR2_P
= STR2, RPM_P = RPM, CPM_P = CP
M, STR1D_P = 0, STR2D_P = 0, CPM
D_P = 0, RANM_P = RANM is executed to initialize various pointers, and RHY = ROCK, ME in 17-2.
AS = 16, TMP = 120, BEAT = * (TBTM
+ TMP), * MPM2 = f000H, KEY = C (=
0) , PREV_PIT = 24 (= C4) is executed, and 1
In 7-3, MODE = NO R MAL and KEYON = NO are executed to set, and finally in 17-4, NTM = f000.
H, MTM = f000H is executed to initialize the tone type and pitch data memory. Generation Start Processing (FIG. 18) The details of the generation start processing routine are shown in FIG . This routine is a routine that is executed when an input device generation start instruction is given. First (18-1) in order to select the music structure, run-level structure generation of music (Figure 19) the lower structure generation (Figure 20). Next, chord progression generation processing 18-2
Running (Figure 22) fit the chord progression database to the upper structure, and selecting the focus cormorants chord progression to a specified rhythm. Next, the rhythm pattern generation process 18-3 (FIG. 23) is executed to select a rhythm pattern that matches the substructure of music, the specified rhythm, and the specified beat from the rhythm pattern database. Next, in step 18-4, T
= 0 to initialize the rhythm counter , 18-5 * N
The tone type memory is initialized by TM = f000H, the pitch memory is initialized by * MTM = f000H in step 18-6, and the state flag is set to the generation mode by MODE = RMELODY in step 18-7. Generation of Upper Structure (FIG. 19) FIG. 19 shows a detailed flow of the upper structure generation routine.
This routine is executed when there is a generation start instruction from the input device 10 or when the end of chord progression is reached. The purpose of the superstructure generation routine is to generate the superstructure of the music and set the substructure that matches the superstructure. The superstructure is generated by extracting the superstructure data from the superstructure database STR1. More specifically, in step 19-1, post = STR1D_P
× 4 + 3, STR 1D = GetB (* STR1_P, p
ost, 4) to execute superstructure database STR1
The upper structure (sentence structure) data STR 1D is extracted from. Subsequently, in 19-2 to 19-5 (19A), the next superstructure is located for the next superstructure generation routine. That is incremented STR1D_P at 19-2, STR1D_P = 4 (step 19-3), or * STR1D_P = If END (step 19-4), STR1D_P = 0, Increment STR
1_P (step 19-5) is executed. Next 19-6
19-10 (19B), the lower structure sequence that matches the upper structure extracted in 19-1 is located. Individual substructures are extracted from the located substructure sequence here (see FIG. 20). In detail, step 1
9-6 executes STR2_P = STR2_P + 3 and, as a result, * STR2_P = ffffH (step 19-
If 7) is satisfied, STR2_P = STR2 (step 19-8) is executed, and GetB is executed in the attribute test 19-9.
It is checked whether (* STR2_P, STR 1D , 1) = 1 holds. GetB (* STR2_P, STR 1
If D 1,1) = 1 is satisfied, STR2D_P = 0 (step 19-10) is executed. As shown in FIG.
Whether eight substructures columns entry into the substructure database STR 2 (8 bars structure) fits to the selected upper structure compares the attribute data of the sub-structure entry and selecting upper structure data STR 1D (eg AA) It is judged by that.

Lower Structure Generation (FIG. 20) FIG. 20 shows a detailed flow of the lower structure generation routine. This routine is called following the superstructure generation routine in the first step 18-1 of the generation start processing.
Further, the current time is called in the step 16-1 3 when barline time interrupt routine (FIG. 16). The purpose of the substructure generation routine is to obtain new substructure data (new bar structure data). More specifically, in step 20-1, post = (STR2D
_P × 4 + 3) mod16, STR 2D = GetB (*
(STR2_P + STR2_P / 4 + 1), post,
4) is executed to obtain the substructure. 20-2 to 20-
4 (20A), the next substructure is located for the next execution of the substructure generation routine. That is, in step 20-2, the lower structural element counter STR2D_P
If (STR2D_P = 8 (step 20-3)) is satisfied by incrementing (pointer to the bar structure in the 8 bar section), STR2D_P = 0 (step 20-4) is executed. GetB (FIG. 21) FIG. 21 is a detailed flow of the function instruction GetB. The function instruction GetB (data, post, nbit) is a function for obtaining a desired partial data item in one word. That is, GetB (data, post, nbit) is 16
This is a function instruction that takes n bits of data to the right of the post bit of one word data of bits, and is called and executed in various routines such as step 19-1 of the upper structure generation routine already described. First, in step 21-1, x1 = data is executed, and then step 21-
2 shifts X1 to the right (post + 1-nbit),
In step 21-3, x2 = ffffh is executed, and 21−
In step 4, X2 is shifted to the left by n bit, in step 21-5, X2 is inverted, and in step 21-6, the bitwise logical product of X1 and X2 is calculated. In the operation example of FIG. 21, 2 bits (01) are extracted to the right from the 8th bit of 1-word data of 16 bits. As a result of the execution of the GetB function instruction, the two bits of 01 are fetched into the 1st bit and 0th (least significant) bit position of the 16-bit word. 22. Chord progression generation routine (FIG. 22) The details of the chord progression generation routine are shown in FIG. This routine step 18-2, or one of the chord progression of generation start process is called in step 16-1 1 each time ends. In the chord progression generation routine, the specified rhythm type is selected from the chord progression database memory CPM ,
And the chord progression that matches the upper structure is taken out. More specifically, in Step 22-1, GetB (* CPM_
P, RHY, 1) = 1 check to the chord progression that access is performed whether or not the attribute test fit the specified rhythm species, G at step 22-2 if someone in this attribute test
etB ((* (CPM_P + 1), STR 1D , 1) =
Check the 1, the chord progression is determined whether it fits in the upper structure. 22-3 to 22-6 when the attribute test of either 22-1 or 22-2 fails.
In, the next chord progression is located from the chord progression database memory CPM. That is, incrementing the word pointer CPM_P chord progression data memory 22-3, at 22-4 * CPM_P = ffffH
It is checked whether or not (the end of the chord progression database), and * (CPM_P-1) = 0H in step 22-6.
It is checked whether or not (beginning of chord progression entry), and if * (CPM_P-1) = 0H is not established, 22-3
Returning to step 22-4, if * CPM_P = ffffH is established, the pointer CPM_P is returned to the head CPM of the chord progression data memory and the process returns to step 22-1, and * (CPM_P-1) = 0H is established in step 22-6. If so, the process similarly returns to step 22-1. Steps 22-1 and 22
Pointer CPM when both attribute tests of -2 are passed
_P locates the proper chord progression. Therefore,
So that the chord pointer points to the first chord in the chord progression
Set to CP_C = 0 in step 22-7 ( code advance
The line length counter is reset to 0 ).

Rhythm Pattern Generation Routine (FIG. 23) FIG. 23 shows a detailed flow of the rhythm pattern generation routine. The rhythm pattern generation routine is step 18-3 of the generation start processing, and step 16-1 for the bar line time.
Called out in 3. The purpose of the rhythm terpolymers emissions generation routine is fit to the designated rhythm species from the rhythm pattern database memory RPM, fit the substructure, is to take out the rhythm pattern (one measure rhythm pattern) and fit the tempo or beat species. The extracted rhythm pattern defines the rhythm pattern (sound length sequence) of the automatically generated one-measure melody. More specifically, in Step 23-1, GetB (* RPM_P, RHY, 1)
By checking whether or not = 1 holds, it is checked whether or not the accessed rhythm pattern matches the designated rhythm type. In Step 23-2, GetB (* (RPM_P +
By checking 1), STR 2D , 1), it is checked whether the rhythm pattern matches the substructure.
23-3 and 23-4 (23A) check whether the rhythm pattern matches the rhythm type corresponding to the tempo.
That is, in 23-3, data = * (RPM_P + 1), p
ost = * (TBTM + TMP) is executed, and it is checked at 23-4 whether GetB (data, 15,4) = post holds. 23-1, 23-2, 23A
23B (23
-5 to 23-7), the next rhythm pattern in the rhythm pattern database is located. That is, 23
Execute RPM_P = RPM_P + 4 at -5 , 23-6
When * RPM_P = ffffH is established and the end of the rhythm pattern database is detected, the RP is performed at 23-7.
Execute M_P = RPM and execute rhythm pattern pointer RP
M_P is returned to the head RPM of the rhythm pattern database. All attribute tests (23-1, 23-2, 2
3A) pass, go to step 23-8 , US
ER_MEL = NO is executed to cancel the melody pattern rule database expansion mode. Stop Processing Routine (FIG. 24) FIG . 24 shows details of the stop processing 15-6 performed in response to the stop instruction 15-5. MOD as shown in 24-1
Execute E = NORMAL to change the system state to normal mode (normal state where melody is not automatically generated)
Have returned to. 25. Key-depression processing routine (FIG. 25) FIG. 25 shows details of the key-depression processing 15-8 executed in response to the key-depression 15-7. 25-1 executes KEYON = YES to set the key depression flag, and 25-2 displays KEYBUF
= KC, PIT = KC is executed , and PIT is pronounced at 25-3. In this way, the sound generation processing of the sound source 12 for the key depression of the keyboard 8 is performed in the key depression processing step 15-8 in the main routine (FIG. 15).

Key Release Processing Routine (FIG. 26) FIG. 26 shows details of the key release processing step 15-10 performed for the key release 15-9 on the keyboard 8. 26-1 K
Set key depression flag key released state running EYON = NO, to mute the PIT at 26-2. Processing routine during key depression (FIG. 27) FIG. 27 shows the details of the processing routine during key depression. This routine is a routine called when 16-2 during key depression (KEYON = YES) is detected in the interrupt routine (FIG. 16). In the processing routine during key depression
Melody pattern rule-based (with the key pressed in the measure as a cue ) is set and the user input melody is patterned in the extended mode (work of creating tone series and pitch sequences). Specifically, 27-1 MO
If DE = NORMAL is established, the program immediately jumps to 27-12 and executes PREV_PIT = PIT. MO
If DE = NORMAL (step 27-1) are not satisfied, i.e., if the mode is the production mode (RMELODY), the flow proceeds to step 27-2, whether the newly depressed key is generated by either i.e. keyboard 8 if PKEYON = NO I see. If the key is continuously pressed, the process jumps to step 27-12. When a new key depression occurs , the routine proceeds to step 27-3, where it is checked whether or not USER_MEL = YES , that is , whether or not it is already in the extension mode. If the already extended mode, jumps to step 27-7. If there is in extended mode, USER proceeds to the step 27-4
_MEL = YES is executed and the fact that there is a melody input from the user in the measure is recorded. Then at 27-5 *
NTM = f000H, * MTM = f000H is executed to initialize the memory of the tone type and the pitch sequence. Then, the mute command is executed at 27-6 to mute the automatically generated melody sound being sounded. 27-3 to 27-6 are extended mode setting processing 27A
Determine. In 27-7, the chord information is created (FIG. 28) to obtain the current chord information, and in 27-8, the sound type classification (FIG.
9) is performed to classify the sound type of the melody sound that is pressed. The results classified in 27-9 are stored in the tone type string memory NTM (tone type store in FIG. 30). Running pitch classification (Fig. 31) classifies the pitch from the preceding tone melody sound according to the key depression at 27-10, and stores the result in pitch column <br/> memory MTM at 27-11 (Fig. 32 pitch store). After that, the program proceeds to 27-12 to execute PREV_PIT = PIT to transfer the contents of the current pitch register to the previous pitch register.

Code Information Creating Routine (FIG. 28) FIG. 28 shows a detailed flow of the code information creating routine.
This routine is step 2 of the processing routine (FIG. 27) during key depression in order to pattern the melody input by the user.
In addition to being called in step 7-7, it is called in step 33-6 of the processing routine (FIG. 33) of the onset time to be described later for automatic melody generation. The purpose of the chord information creation routine is to retrieve the chord information at the current time from the chord progression selected from the chord progression database. More specifically, the code length accumulator is initialized by i = 0 at 28-1 and j = at 28-2.
The head code of the selected code progress data is located by CPM_P + 2. Next, 28A (28-3 to 28-5)
Locate the current code (code at the current time). That is, in 28-3, i = i + GetB (* j,
5 and 6) is executed, and at 28-4, it is checked whether i> CP_C. If not satisfied, J is incremented at 28-5. After locating the current code, the process proceeds to step 28-6 to fetch the data of the current code into the CHO register by CHO = * j. Then at 28-7, CHO (ROOT) =
(GetB (CHO, 15,4) + KEY) mod1
2, CHO (TYPE) = GetB (CHO, 11,
6) Obtain the root and type information of the current code by executing 6). Sound Type Classification Routine (FIG. 29) The details of the sound type classification routine are shown in FIG. This routine is step 2 for patterning the user-entered melody.
It is called in step 7-8 and is called in step 33-7 of the processing routine of the sounding time (FIG. 33) which will be described later in order to generate an automatic melody. The purpose of the sound type classification routine is to classify the sound type of the melody sound of interest (a melody sound on a key depression or a sound candidate for an automatically generated melody sound). The chord information, the key information KEY , the designated rhythm type information RHY and the standard pitch class set of chord tone , scale note, and tension note from the chord information creation routine (FIG. 28) for classifying such note types are used. In detail, step 2
9-1 pos1 = (12 + PIT-CHO (ROO
T)) mod12, pos2 = (12 + PIT-KE
Y) Execute mod12. POS1 is pitch PI here
It represents the pitch from the chord root of T, and POS2 is the pitch P.
Indicates the pitch from the IT key. Next, at 29-2, x1 =
PCS (CT) = * (CT + CHO (TYPE)), x
2 = PCS (TN) = * (TN + CHO (TYP
E)), x3 = PCS (SN) = * (SN + RHY) to obtain each pitch class set (PCS) of chord, tension and scale. Next, at 29-3, GetB (x
1, pos1, 1) = 1 (code tone) is checked, and if satisfied, NT = CHOT is executed at 29-4 . NT = CHOT declares that the melody sound having the pitch PIT is a chord tone. In Step 29-5, GetB (x2, pos1,1) = 1a
Whether or not ndGetB (x3, pos2,1) = 1 is satisfied, that is, whether or not it is an available note , is checked , and if so, NT = AVAI is executed at 29-6. NT
= AVAI is a declaration that the focused melody sound is an available note. Step 29
At -7, GetB (x3, pos2,1) = 1 is used to check whether the sound of interest is a scale note, and if so, to declare that fact, NT = SCA at 29-8.
Execute L. In Step 29-9, GetB (x2,
It is checked whether the melody sound is a tension note by pos1, 1) = 1, and if satisfied, NT = TENS is executed at 29-10 to declare that. 29-
If any of 3, 29-5, 29-7, and 29-9 is not established, 29-11 declares that the melody sound is an void note by NT = AVOI. Reference number 29
In the operation example of 0, the PIT is F when the code is D MAJOR.
It shows how the sound of # 3 is classified as a chord tone. Ben (Venn) circle X1 as shown in FIG. 291 code tone PCS (CT), the circle X2 is tension note PCS (TN), the circle X3 is a scale note PC
S (SN) . The melody sound determined to be an element of X1 is a chord tone. The portion where the sets X2 and X3 overlap is determined as an available note. Not overlap with the set X1 of the set X3, and does not overlap with the set X2 moiety is a region that is determined to scale note. Set X2
A portion that does not overlap with the set X3 is an area determined as a tension note. The area outside the gathering circles X1, X2, and X3 is an area determined as an void note.

Sound Type Store Routine (FIG. 30) FIG. 30 shows the details of the sound type store routine. This routine patterns the user-input melody (step 27).
- 9) and Invoked in the process for the evaluation of automatic melody tone (in step 33-8). In the tone type store routine, the tone type string N is stored in the NTM using the tone type string memory NTM as a right shift register . That is, 30-1 shifts NTM right by 4 bits , 30-2 shifts NT left by 12 bits , and 30-3 shifts the logical sum of NT and NTM by N.
Substituted in TM. Pitch Classification Routine (FIG. 31) FIG. 31 shows details of the pitch classification routine. Pitch classification routine is Invoked in the step 27-1 0 and 33-7. In the pitch classification routine, the current pitch PIT is changed to the previous pitch P.
Compared to REV_PIT, it is classified pitch formed between Ryooto the (motion). That is, PIT = PREV
If _PIT (step 31-1), MT = SAME (step 31-2) is declared to indicate no motion , and PI
If T-PREV_PIT> 2 (step 31-3), M
T = + JUMP (step 31-4) declares the upward jumping progress, and if PREV_PIT-PIT> 2 (step 31-5), MT = -JUMP (step 31-
6) Declare the downward jump progress, and PIT-PRE
If V_PIT> 0 (step 31-7), MT = + ST
EP successively progress declared upward (step 31-8), PREV_PIT-PIT> 0 ( step 31-
If it is 9), MT = -STEP (step 31-10) is used to declare downward sequential progress. Pitch Store Routine (FIG. 32) FIG. 32 shows the details of the pitch store routine. This routine Invoked in Step 27-1 1 (melody pattern rule base extension system) and the step 33-8 (Automatic melody tone generating system). The pitch store routine operates the pitch string memory MTM as a right shift register to store the pitch data M T which is the classification result in M TM. Ie 3
2 -1 shifts M TM to the right by 4 bits , and 3 2 2 shifts M T
The 12-bit left shift, and substitutes a logical OR of M T and M TM 3 2 -3 M TM.

Sounding Time Processing Routine (FIG. 33) FIG. 33 shows the details of the sounding time processing routine. This routine is a routine that is executed at the time of the onset time of the auto-generated melody tone (16 6). That is, this routine is called when the current time is the timing of bit "1" in the key-on pattern of the rhythm pattern. The purpose of the processing routine for the pronunciation time is to determine the pitch of the melody tone and generate it. For this reason to make a pitch candidate in the processing routine of the onset time by adding a random number from the random number memory RANM to the pitch of the previous melody tone current melody tone, to inspect the melody pattern of up to the candidate by the melody pattern rule-based. If the rule base has an entry that matches the pattern up to the current pitch candidate , the current melody candidate is determined as the pitch of the current note. If there is no entry corresponding to the melody pattern rule base, another pitch candidate is created and the inspection is repeated by the rule base. More specifically, in 33-1, the tone pitch interval memory NT
Right shift M and MTN by 4 bits. 33-2 generates a pitch candidate random Markov. That is, the random number * RANM_P from the random number memory is added to the pitch of the preceding sound PREV_P.
IT is added to generate a current sound candidate. Next, 33A (33
-3 to 33-5), locate the next random number for the next routine execution. That is, the random number pointer RANM_P is incremented in 33-3 and * RA is calculated in 33-4.
When the end of the random number memory is detected by NM_P = fffH, the random number pointer is returned to the head of the random number memory by RANM_P = RANM (step 33-5). 33-6
Call the code information creation routine with. Loop 33-7
~ 33-12 at the entrance 33-7 of the sound type classification routine (Fig. 2
9) and calls the pitch classification routine (FIG. 31), classifies the note type and pitch for the pitch candidate PIT. 33-
Call 8 sound type store routine (FIG. 30) the pitch Store routine (FIG. 32), loads the classification result of the sound type and pitch note type column memory NTM respectively, as a right end of the nibble in the pitch column memory MTM. Call the evaluation routine (FIG. 34) at step 33-9, the melody pattern to the current sound pitch candidate PIT defined by NTM and MTM is to check which melody pattern rule base. Without the rule base is entry over N
It becomes G, and the next pitch candidate is generated in 33B (33-10 to 33-12). That is, n is incremented at 33-10 executes Z = -1 × Z at 33-12.
Then, the process returns to step 33-7 and the process is repeated for the pitch candidate generated in 33B. 33B successively generates pitch candidates such as the first pitch candidate ± 1, ± 2, ± 3 obtained in 33-2 each time (here, +1 is a semitone and -1 is a semitone). Lower, +2 is 2 semitones up, -2 is 2 semitones down ...). Become a GOOD if the entry in the evaluation routine 33-9 to match the Merodipa Tar down until the sound pitch candidate is found from the rule base, PREV
_PIT = PIT (step 33-13) is executed and 3
In 3-14, the key-on flag is set to ON by KEYON = YES, and the determined pitch PIT is sounded. Thus there is a sign of key-on from the key-on patterns in the automatic melody generation mode, melody sound is determined in accordance with the melody pattern rule base rules, it is pronounced.

Evaluation Routine (FIG. 34) FIG. 34 shows the details of the evaluation routine called in step 33-9 of FIG. The purpose of the evaluation routine is the tone series N
It is to determine whether the melody pattern defined by TM and the pitch sequence MTM follows the melody pattern rule (determines whether it matches the rule entered in the melody pattern rule base). Here, the melody pattern rule base is a fixed melody pattern rule base MPM1 and RAM placed in the ROM4.
Extended melody pattern rule base MP formed in 6
There is M2. The evaluation routine searches both routines for a rule that matches the NTM and MTM patterns. 34A (34-1 to 34-4) is a fixed melody pattern rule-based search, and 34B (34-5).
34-8) is a search for the extended melody pattern rule base. In 34-1, the head of the fixed melody pattern rule base is located by i = MPM1 and the loop 34
At the entrance of -2 to 34-4, * i = NTMand * (i +
1) = MTM (step 34-2) , it is checked whether the tone type sequence and the tone sequence of the accessed rule match N T M and M T M, respectively. If they do not match, i = i + 2 (step 34-3) to locate the next rule, fixed melody pattern rule base end at step 34-4 (GetB (* i, 15,4 ) =
Until END) is not detected, the processes of 34-2 and below are repeated. In step 34-2, (* i = NTMand *
If (i + 1) = MTM) holds, GOOD is returned, 3
Locate the beginning of the extended melody pattern rule base by 4-4 in a fixed melody pattern rule-based 34-5 with i = MPM2 After termination is detected, the loop 34
At the entrance (34-6) of -6 to 34-8, it is checked whether the tone type sequence and the tone sequence of the accessed rule match NTM and MTM, respectively. If they do not match, the next rule is located by i = i + 2 (step 34-7) and the end of the extended melody pattern rule base (GetB (* i, 15,4) = END) must be detected in step 34-8. If the condition of 34-6 is satisfied, GOOD is returned. If the end of the extended melody pattern rule base is reached at 34-8, NG is returned. Mute Time Processing Routine (FIG. 35) The details of the mute time processing routine called in step 16-8 at the time of mute of the automatic melody sound (the time when the bit "1" representing the key off is given from the key off pattern) are described. It shows in FIG. If KEYON = YES (step 35-1), the sound PIT is muted because it is sounding ,
Set the key-on flag to KEYON = NO (35-
2). 36. Melody pattern DB expansion routine (FIG. 36) FIG. 36 shows the details of the melody pattern database expansion routine called in steps 16-13 at the bar line time. The purpose of this routine is to add the melody pattern input by the user (the melody pattern input during one bar) to the melody pattern rule base MPM2 of the RAM 6. More specifically, if USER_MEL = YES (step 36-1) is not satisfied , there is no melody input from the user during one bar , and the process returns . USER_M
When EL = YES (36-1) is satisfied , 36A (3
6-2 to 36-7), the end of the extended melody pattern rule base MPM2 is detected. That is, i = MPM2
(36-2) locates the pointer to the beginning of the extended melody pattern rule base, and * i at the entrance of the loop.
= FfffH (36-3) is not established, i is incremented (36-6), and the memory is over (i
> MPM2 + MPM2SIZE), if * i = ffffH (36-3) holds, * i =
* NTM, * (i + 1) = * MTM, * (i + 2) = f
fffH (step 36-4) is executed to set the user melody pattern to the melody pattern database MPM2.
Write to * NTM = f000H, * MTM = f000
H (step 36-5) re-initializes the tone type and pitch sequence memory.

Pattern data format adjustment (see FIG. 3)
7 to 39) The flow of FIGS. 37 to 39 relates to the format adjustment between the NTM and MTM and the melody pattern rule bases MPM1 and MPM2. The routine of FIG. 37 is the first step (34 of FIG. 34) of the evaluation routine (FIG. 34).
-1) and the melody pattern rule base expansion routine (FIG. 36) are called in step 36-4 (see FIG. 39, which is a detailed flow of step 36-4). The routine of FIG. 37 is for adjusting the format of the melody pattern to be input or evaluated by NTM and MTM to the format of the rules of the rule bases MPM1 and MPM2. In the case of the embodiment, the MTM has only one extra pitch data item from the past melody tone type discarded on the NTM at the input of the evaluation routine or the input of the melody pattern rule base extension routine. In order to match this with the pitch sequence format of the rule-based rule, it is necessary to convert this extra pitch data item (nibble) into an END nibble. This processing is performed by the routine of FIG. In the figure, the symbol "∨" represents the bitwise logical sum, and "∧" represents the bitwise logical product. The routine of FIG. 38 is the evaluation of steps 34-2 and 34-6 of the evaluation routine (FIG. 34). In this routine (FIG. 38), even if not the entire melody pattern indicated by the rules from the rule bases MPM1 and MPM2,
The partial pattern is the melody pattern (NT
If it matches with M and MTM), the melody pattern to be evaluated is processed according to the rule (returning YES) (this is effective in saving the number of rules in the rule base). is there). FIG. 39 shows details of step 36-4 of the melody pattern database expansion routine. The above description, the description of FIGS. 37 to 39 itself,
37 to 3 from the description of other drawings and the description in the specification.
Since the details of the routine of No. 9 are clear, further explanation is omitted.

[0030]

[Modification] Although the description of the embodiment has been completed, various modifications and changes can be made within the scope of the present invention. For example, the melody pattern rule base 216 may have a melody pattern (MP) rule base for each music style. FIG. 40 shows a melody pattern rule base for each music style and its related configuration. The melody pattern rule base has a rule base for each of a plurality (three in FIG. 40) of music styles. In FIG. 40, 216A is a style No. 1-No. 3 represents an MP rule group common to all styles (which can be used in any style). 1 and No. 2 represents a common MP rule group, and 216C is a style number. 1 represents an MP rule group unique to 1
6D is style No. 2 represents the MP rule group unique to the item 2. Style No. There is no MP rule group specific to 3). Therefore, the style No. The melody pattern rule base of 1 is 216A, 216B, and 216.
Style No. The melody pattern rule base of No. 2 is defined by 216A, 216B, and 216D. The melody pattern rule base of 3 is defined by 216A. In the operation of the automatic composer, the selection module 230 creates a selection MP rule base 232 for the designated music style (selects from the rule base 216) in accordance with the musical style input (for example, the designated rhythm 131 in FIG. 1). The selection MP rule base 232 is the melody pitch string generation device 200 of FIG. 1B.
Is accessed by an evaluation module, such as Matching 214 of. By using the configuration as shown in FIG. 40, the automatic composer can more efficiently generate a melody that matches the musical style. It is also easy to speed up data retrieval from the database to support full real-time response of composition. 41 and 4 for an example.
2 shows. FIG. 41 shows a modified example 400 of the music material database and its search method. In FIG. 41, the database of music materials (for example, rhythm pattern, chord progression) includes an index table 403 and a material database body 405. It is assumed that the first attribute (for example, music structure) is given as “α” and the second attribute (for example, musical style) is given as “β” as the composition condition as shown in 401 (the number of attribute items is not limited to 2, though). For convenience of illustration, the number of attribute items is 2.) It is assumed that there are M types of first attributes and N types of second attributes. The index table 403 gives index information to the material database main body 405, and for each composition condition (combination of attributes), the storage location of the material which is placed in the material database main body 405 and matches the composition condition (start address ) And the number of entries for that material. In operation, from composition condition 4-1 to block 4
02, INDEX + (N × α + β) × 2 is calculated, the index table 403 is addressed, and the address X to the material database body 405 written in the index table 403 (for example, if the first attribute is “1” 2 Read the starting address X1) of the material entry group whose attribute is "1" and the number of entries S (for example, S1) written at the next address. Next, as indicated by 411 in the search module 410, 0
Generate a random number RAN of ~ 1, and generate address X and number of entries S
And the number of words / entry in the database main body 405, for example, 3 is used, 412 to 414 (RA
N * S) * 3 + X is calculated to generate the address ADDR of the music material entry to be searched. This address ADD
R (and the number of consecutive words / number of consecutive addresses)
By accessing the music material database main body 405, it is possible to search the database for a desired music material, that is, a music material satisfying the composition condition 401, and generate it as a part of composition.

According to the configuration shown in FIG. 41, desired data retrieval with respect to the music material database 405 can be performed very quickly. However, the database body 405 contains duplicated data, and the memory efficiency is not good. FIG. 42 shows another modification 500 of the music material database and its search method. In this modified example 500, the music material database has first and second index tables 503, 50.
4 and a database body 505. In this case, the data does not overlap on the database body 405. The second index table 50 is stored in the first index table 503.
The index (address) information for No. 4 is written. The first index table 503 can be immediately accessed from the composition condition (first attribute = α, second attribute = β).
That is, as shown by 502, the table 503 can be immediately addressed by the calculation of INDEX = 1 (the start address of the first index table 503) + (N × α + β). As indicated by 511, the first index table 503.
The second index table 504 is accessed with the data X read from. In the second index table 504, the number S of material entries in the database main body 505 (for example, S1 for conditions 1 and 1) and the address list of each material entry are written as composition conditions. Therefore, 51
As shown in FIG. 2, the (RAN × S) th item (where RAN is a random number from 0 to 1) in the address list is read, and the material database main body 505 is accessed to immediately find the material that meets the desired composition condition 401. You can take it out. According to the configuration of FIG. 42, the database main body 5
It is possible to quickly generate a music material (for example, melody rhythm, chord progression) that meets the set composition condition while avoiding data duplication in 05. In addition, various modifications and applications are possible.

[0032]

As described above in detail, according to the present invention, it is possible to adapt to music progression through devising musical knowledge expression (use of melody pattern rule base), utilizing music progression, generating melody, etc. Moreover, a good melody supported by music knowledge can be composed in real time or in an environment close to real time.

[Brief description of drawings]

FIG. 1A is a functional block diagram of a composition conditioner incorporated in an automatic composer according to the present invention.

FIG. 1B is a functional block diagram of a real-time melody pitch string generation device incorporated in an automatic composer.

FIG. 1C is a functional block diagram of a melody pattern rule-based expansion device incorporated in an automatic composer.

FIG. 2 is a block diagram of typical hardware that realizes an automatic composer.

FIG. 3 is a functional block diagram of an automatic melody generator incorporated in the automatic composer of FIG.

FIG. 4 is a diagram showing character (symbol) expressions and numerical expressions of music elements used in the embodiment.

FIG. 5 is a diagram showing character and numerical expressions of still another music element.

FIG. 6 is a diagram showing character and numerical expressions of still another music element.

FIG. 7 is a diagram showing a standard pitch class set memory.

FIG. 8 is a view showing a fixed melody pattern database memory and a tone type / pitch data memory.

FIG. 9 is a diagram showing a tempo / bit conversion map and a random number memory.

FIG. 10 shows a chord progression database memory.

FIG. 11 is a diagram showing a music structure database memory.

FIG. 12 is a view showing a rhythm pattern database memory.

FIG. 13 is a diagram illustrating other constants and variables.

FIG. 14 is a diagram for explaining other variables.

15 is a flowchart of a main routine executed by the CPU of FIG.

16 is a flowchart of an interrupt routine executed by the CPU of FIG.

FIG. 17 is a detailed flowchart of initialization.

FIG. 18 is a flowchart of a generation start processing routine.

FIG. 19 is a detailed flowchart of an upper structure generation routine.

FIG. 20 is a detailed flowchart of a lower structure generation routine.

FIG. 21 is a diagram showing a detailed flowchart of GetB and an operation example.

FIG. 22 is a detailed flowchart of a chord progression generation routine.

FIG. 23 is a detailed flowchart of a rhythm pattern generation routine.

FIG. 24 is a flowchart of stop processing.

FIG. 25 is a flowchart of key depression processing.

FIG. 26 is a flowchart of key release processing.

FIG. 27 is a flowchart of processing during key depression.

FIG. 28 is a detailed flowchart of a code information creation routine.

FIG. 29 is a detailed flowchart of sound type classification.

FIG. 30 is a detailed flowchart of the sound type store.

FIG. 31 is a detailed flowchart of pitch classification.

FIG. 32 is a detailed flowchart of pitch store.

FIG. 33 is a detailed flowchart of processing of pronunciation time.

FIG. 34 is a detailed flowchart of an evaluation routine.

FIG. 35 is a flowchart of processing of muffling time.

FIG. 36 is a detailed flowchart of a melody pattern database expansion routine.

FIG. 37 is a detailed flowchart of MTM format change.

FIG. 38 is a detailed flowchart of steps 34-2 and 34-6.

FIG. 39 is a detailed flowchart of step 36-4.

FIG. 40 is a view showing a configuration related to a melody pattern rule base for each music style.

FIG. 41 is a diagram showing a modification of the music material database and its search method.

FIG. 42 is a diagram showing still another modified example of the music material database and its search method.

[Explanation of symbols]

200 Melody Pitch Sequence Generator 216 Melody Pattern Rule Base (Melody Pattern Rule Base Means) 132 Specified Beat or Tempo (Tempo Designating Means) 228 Real Time Performance System (Real Time Melody Performance Means) 302 User Input Melody (User Melody Input Means) 310 melody pattern (tone type / tone sequence) generation block (melody pattern recognition means) 330E melody pattern rule base extension unit 208 tone type / tone sequence generation module (melody pattern storage means, tone / tone type classification means, evaluation target melody pattern) Forming means) 214 Matching module (rule-based searching means) 220 Next candidate generating module (different candidate generating means) 208, 214, 218, 220 (repeating means) 224 Determined current pitch (pitch determining means) 110,116,117,120,140,145,1
60 (chord progression generation means) 210 current tonality (tonality indication means)

─────────────────────────────────────────────────── ─── Continuation of front page (58) Fields investigated (Int.Cl. ⁷ , DB name) G10H 1/00-7/12 G09B 15/00-G10B 15/08 G10G 1/00-7/02

Claims (4)

(57) [Claims] 1. A music progression giving means for giving a music progression, a melody pattern rule base means for storing a rule base of a melody pattern in which a tone series of a melody is patterned by a tone series and a tone series, and a melody input by a user. Melody input means
And the tone type of each sound of the melody input by the melody input means
To the music progress given by the above-mentioned music progress giving means.
Judgment based on line and sound of adjacent melody
By determining the pitch, it is represented by a pitch sequence and a pitch sequence.
Melody pattern creator
And a melody pattern created by the melody pattern creating means.
The above melody pattern rule as an additional rule
Expansion of melody pattern rule base to be added to bass means
An automatic composer comprising: an extension means; and a melody generation means adapted to generate a melody adapted to the music progression from the music progression giving means and satisfying the melody pattern rule in the melody pattern rule base means. . 2. Automatic music for automatically generating melody sounds in sequence
In the machine, a music progression giving means for giving a music progression and a method in which the tone sequence of the melody is patterned into a tone sequence and a tone sequence.
A melody pattern that memorizes the rule base of the roddy pattern
Stored in the melody pattern rule base means and the melody pattern rule base means
Without referring to the melody pattern rule base, the current
Melody sound candidate generator for generating existing melody sound candidates
And the current melody sound generated by the melody sound candidate generation means.
The candidate sound type is added by the music progression adding means.
Sound type determination means for determining based on the progress of music, and the current melody generated by the melody sound candidate generation means.
The pitch between the sound candidate and the immediately preceding melody sound that has already been generated
The pitch determination means, the pitch type determined by the pitch type determination means, and the pitch determination hand.
The sequence of past note types and pitch sequences that includes the
The turn is recorded in the above melody pattern rule base means.
Rule-based search means to determine whether or not it is remembered
And it is determined that the rule-based search means stores it.
Sometimes, the current melody sound candidate is used as the current melody sound.
And the next melody following the melody sound
Instructing the melody sound candidate generation means to generate sound candidates
However, it is determined that it is not stored by the above rule-based search means.
When rejected, a new alternative to the current melody candidate above
Generating a melody sound candidate
Automatic composer, characterized in that it comprises an instruction means for instructing, to the. 3. A automatic composer according to claim 2, melody input means capable melody input by user, the sound type <br/> each sound of the melody inputted by the melody input means, the Music progression provided by the music progression providing means
Judgment based on line and sound of adjacent melody
By determining the extent, and the melody pattern created hand <br/> stage to create a melody pattern represented by a note type column and pitch sequence, the melody pattern generated by the melody pattern creating means, as additional rules An automatic composer, further comprising: a melody pattern rule base expansion means added to the melody pattern rule base means. 4. The automatic composer according to claim 2, wherein the melody sound candidate generation means is the music progress addition means.
Without referring to the music progression given in
An automatic composer characterized in that it generates candidates for a rody sound .