JP2000122664A - Vibrato generating device and method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To generate a vibrato sound of high quality with a low storing capacity. SOLUTION: Plural waveform data taken out dispersedly from an original waveform A imparted with a vibrato are stored, one of the wave form data is read out repeatedly with prescribed intervals and the wave form data to be read out is switched in order to conduct a prescribed reading-out sequence, and the reading-out sequence is repeated to realize a vibrato ranging over plural periods. Plural dispersed waveform data are stored corresponding to the plural periods of the vibrato, vibrato priodic positions to which the respective waveform data belong are changed, while keeping relative time position within one period of the vibrato of each waveform data, to optionally variably set an order for reading out respective data ranging over the plural periods of the vibrato, the waveform data is repeatedly read out in a prescribed interval, and the waveform data to be read out are switched in order according to a set read-out order, so as to realize the vibrato ranging over the plural periods.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a vibrato generating apparatus and method, and more particularly to a technique capable of synthesizing a high-quality musical sound waveform having articulation, and is not limited to electronic musical instruments. The present invention can be widely applied as a vibrato generator and / or method for a musical sound or sound generator for various uses such as a game machine, a personal computer, and other multimedia devices. In this specification, "musical sound"
The term is not limited to the sound of music, but is used in a broad sense including general sounds, such as human voices, various sound effects, and sounds in the natural world.

[0002]

2. Description of the Related Art In a sound source of a waveform memory readout system (PCM: pulse code modulation system) used in electronic musical instruments and the like, one or a plurality of cycles of waveform data corresponding to a predetermined tone color are stored in a memory. A continuous tone waveform is generated by repeatedly reading the waveform data at a desired reading speed corresponding to a desired pitch (pitch) of the tone to be generated. Also, by storing data of all the waveforms from the start to the end of the tone generation in the memory and reading out the waveform data at a desired read speed corresponding to a desired pitch (pitch) of the tone to be generated. One tone is also generated. In this type of PCM tone generator, when a waveform stored in a memory is simply read out as it is to generate not only a musical tone but also a certain change to make the generated musical tone expressive, the pitch, volume, etc. , And tone colors. As for the pitch, by modulating the readout speed appropriately according to an arbitrary pitch envelope, a pitch modulation effect such as vibrato or attack pitch is provided. Regarding the volume, it is possible to add a volume amplitude envelope according to a required envelope waveform to the read waveform data, or to apply a tremolo effect or the like by periodically modulating the volume amplitude of the read waveform data. Done. As for the timbre, appropriate timbre control is performed by filtering the read waveform data.

In addition, continuous performance sounds (phrases) actually played live are collectively sampled and pasted (recorded) on one recording track, and the phrase waveforms pasted on a plurality of tracks are collected. There is also known a multi-track sequencer that reproduces and sounds in combination with an automatic performance sound based on separately recorded sequence performance data. Also, a method of recording all musical tone waveform data of one tune actually played live as PCM data and simply reproducing the data is well known as a music recording method for a CD (compact disc).

[0004]

By the way, when a skilled performer of an arbitrary natural musical instrument such as a piano, a violin, a saxophone or the like plays a series of musical phrases with the musical instrument, the contents of the performance sounds are the same. Despite being played on musical instruments, it is not uniform,
Alternatively, in the connection between sounds, or at the rising, sustaining, or falling parts of a sound, the music is played with a slightly different "articulation" depending on the musical composition or the sensitivity of the performer. You. The existence of such "articulations" gives the listener the impression of a really good sound. The method of recording all the music performances performed by a skilled player as PCM waveform data, like the music recording method on a CD, is capable of real and high-quality reproduction of a live performance. You can realistically reproduce street articulations. However, since it can be used only as a mere reproduction device for a fixed song (song as recorded), an interactive musical instrument or multimedia device that allows the user to freely create and edit sounds. It cannot be used as music creation technology.

On the other hand, a PCM known in electronic musical instruments and the like is used.
As described above, the sound source technology allows a user to make a sound, and allows the generated musical sound to have a certain expressive power. However, it was insufficient to realize natural "articulation" in both sound quality and expressiveness. For example,
In general, in this type of PCM sound source technology, the waveform data stored in the memory merely stores a sample of a single tone played by a natural musical instrument, and thus the sound quality of generated musical tones is limited. In particular, articulation or playing style of connection between sounds during performance cannot be expressed with high quality. For example, in the case of a slur playing method in which a preceding sound is smoothly changed to the next sound, in a conventional electronic musical instrument or the like, the waveform data reading speed from the memory is simply changed smoothly, or the volume given to the generated sound is changed. It merely relies on techniques such as controlling the envelope, and has not been able to achieve articulation or playing techniques with sound quality comparable to live performance of natural musical instruments. In addition, different articulations such as the rising part of the same musical instrument at the same pitch, depending on the difference in the song phrase, or even on the same song phrase, depending on the difference in performance occasions. However, such subtle differences in articulation can also be expressed by a PCM known in electronic musical instruments.
It could not be realized in sound source technology.

Also, the control of generated musical tones according to performance expressions is relatively monotonous in conventional electronic musical instruments and the like, and cannot be said to be sufficient. For example, it is known to perform musical tone control according to a performance touch such as a key, but also in this case, it is only possible to control a volume change characteristic and a tone filter characteristic according to the touch. For example, it has not been possible to freely control the tone characteristics for each partial section of the entire tone generation section from the rise to the fall of a tone. Regarding the tone color control of the generated tone, once one tone color is selected prior to the performance, waveform data corresponding to the selected tone color is read from the memory, and thereafter, various performance expressions are generated during the tone generation. Therefore, the waveform data corresponding to the timbre is only variably controlled by a filter or the like, so that the timbre change according to the performance expression is not sufficient. In addition, the control envelope waveforms such as pitch and volume are set and controlled in terms of the shape, etc., of a series of envelopes from the rise to the fall of the envelope as one unit. It has not been made available.

On the other hand, in a system such as the above-described multi-track sequencer, only the live performance phrase waveform data is pasted. This method cannot be used at all as an interactive musical sound creation technology that allows a user to freely create sounds in electronic musical instruments and multimedia devices.
Not only musical performance sounds, but also general sounds that exist in the natural world are rich in delicate “articulations” according to their lapse of time. The "articulation" of existing sounds could not be controllably reproduced.

Further, the conventional typical vibrato imparting technique is such that the speed at which a musical tone waveform is read from a waveform memory is periodically modulated in accordance with a vibrato frequency, and only a monotonous vibrato sound can be generated. Was. On the other hand, it is conceivable to record the original waveform to which vibrato was originally added in a memory and read and read it, but this has poor controllability during reproduction. In addition, the recording capacity of the memory also increases.

The present invention has been made in view of the above points, and provides a vibrato generating apparatus and method capable of generating a high-quality vibrato sound with good controllability and saving data storage capacity. It is something to offer. In addition, when musical sounds (including not only musical sounds but also other general sounds) are generated using electronic musical instruments and electronic devices, "articulation" is realistically reproduced. It provides interactive high-quality musical sound creation technology that allows the user to freely create and edit sounds in electronic musical instruments and multimedia devices, etc. It is an object of the present invention to provide a vibrato generating apparatus and method capable of efficiently generating a high-quality vibrato sound.

In this specification, the term "articulation" is used in a generally known meaning, and includes, for example, "syllable", "connection between sounds", "several sounds". Lump (phrase) ",
"Partial characteristics of sound", "pronunciation method", "playing method",
It shall be used in a wide concept including all the concepts such as "performance expression".

[0011]

According to the present invention, there is provided a vibrato generating apparatus for storing a plurality of waveform data which are dispersedly extracted from an original waveform to which vibrato is added, and storing one of the waveform data. Reading means for repeatedly reading out data during a predetermined time and sequentially switching waveform data to be read out, thereby executing a predetermined reading sequence, and realizing vibrato over a plurality of cycles by repeating this reading sequence. is there.

With this arrangement, a plurality of waveform data are dispersedly taken out from an appropriate range such as one or more periods of vibrato or less than one period of the original waveform, and stored. It is possible to save the storage capacity of the waveform data stored in the storage means, and to obtain high-quality vibrato waveform data that reflects vibrato characteristics in the original waveform as much as possible by distributed extraction. it can. Then, one of the waveform data is repeatedly read out in a predetermined interval, and the waveform data to be read out is sequentially switched to execute a predetermined readout sequence. By repeating this readout sequence, vibrato over a plurality of cycles is obtained. Can be realized. Since each waveform data extracted in a dispersed manner reflects the characteristics of vibrato in the original waveform, the tone waveform obtained by repeating and combining these waveforms has a high-quality vibrato comparable to the vibrato in the original waveform. Is included. Here, by variously controlling how to read each waveform data from the storage means, it is possible to variously control various vibrato elements such as a vibrato cycle. Therefore, according to the present invention, there is an excellent effect that high-quality vibrato sound can be generated with rich controllability and the data storage capacity can be saved.

In one embodiment, the apparatus further comprises control data generating means for generating control data indicating a temporal change in pitch in response to a read sequence, wherein the read means comprises:
The reading speed of the waveform data may be temporally changed according to the control data. A control data generating means for generating control data indicating a temporal change of the amplitude in accordance with the read sequence; and controlling the amplitude of the waveform data read by the read means temporally in accordance with the control data. An amplitude control means may be further provided. As a result, it is possible to control the vibrato depth and the accompanying amplitude fluctuation. Further, the apparatus further comprises control data generating means for generating time control data for controlling the time of one cycle of vibrato, and the time for repeatedly reading out one waveform data in the reading means is variably controlled by the time control data. The time of the read sequence may be controlled to variably control the vibrato cycle. Further, a cross-fade synthesizing means for cross-fade synthesizing the preceding waveform data and the succeeding waveform data while repeatedly reading each of them at predetermined intervals may be provided.

According to another aspect of the present invention, there is provided a vibrato generating apparatus which stores a plurality of dispersed waveform data corresponding to a plurality of cycles of vibrato, and stores a plurality of pieces of waveform data relative to each other within one vibrato cycle. Setting means for arbitrarily setting the reading order of each waveform data over a plurality of cycles of vibrato by changing the vibrato cycle position to which each waveform data belongs while maintaining the time position; and And a read means for realizing vibrato over a plurality of cycles by repeatedly switching waveform data to be read in the set reading order. As described above, the vibrato sound is synthesized in various combinations without giving a sense of incongruity by replacing the vibrato cycle position to which each waveform data belongs while maintaining the relative time position within one vibrato cycle of each waveform data. be able to.

The art of musical sound data creation and musical sound synthesis according to the present invention analyzes sound articulation and performs musical sound editing and synthesizing processing in units of articulation elements to model sound articulation. To perform musical tone synthesis. Therefore, this technology is called SAEM (Sound Articulation Element).
Modeling) technology. The present invention can be constructed and implemented not only as a method invention, but also as an apparatus invention. Further, the present invention can be implemented in the form of a computer program, or can be implemented in the form of a recording medium storing such a computer program. Furthermore,
The present invention can be embodied in the form of a recording medium storing waveforms or musical sound data having a novel data structure. The technique of generating a vibrato sound proposed in the present invention can be applied not only to high-quality waveforms including articulations but also to musical sound waveforms having other general waveforms.

[0016]

Embodiments of the present invention will be described below in detail with reference to the accompanying drawings. [Example of Creating Musical Tone Database] As described above, when a skilled performer of an arbitrary natural musical instrument such as a piano, violin, or saxophone plays a series of musical phrases with the musical instrument, the contents of the musical sounds are the same, for example. Despite being played on musical instruments, it is not uniform,
Alternatively, in the connection between sounds, or at the rising, sustaining, or falling parts of a sound, the music is played with a slightly different "articulation" depending on the musical composition or the sensitivity of the performer. You. The existence of such "articulations" gives the listener the impression of a really good sound. When playing an instrument,
"Articulation" is a "playing technique" by a performer
Or, it appears as a reflection of "performance expression". Therefore, in the following description, it is to be noted in advance that the terms "playing style" or "performance expression" and "articulation" are sometimes used to indicate substantially the same meaning. For example, "playing techniques" include stakart, tenuto, slur, vibrato, tremolo, crescendo,
There are various other things such as decrescendo. When a performer plays a series of musical phrases with musical instruments,
Various playing styles are used in each performance phase according to the instructions in the score or according to one's own sensibility, producing "articulations" corresponding to each playing style.

FIG. 1 shows an example of a procedure for creating a tone database according to the present invention. First step S1
Is a step of sampling a series of performance sounds composed of one or more musical tones. Here, for example, a skilled performer of a certain natural musical instrument plays a predetermined series of music phrases with the musical instrument. This series of performance sounds is picked up by a microphone, sampled according to a predetermined sampling frequency, and PCM-encoded waveform data for the entire performance phrase is obtained. This waveform data is high quality data excellent in music. For the purpose of explanation, FIG. 2A shows an example of a musical score of a series of music phrases played for sampling in step S1. The “playing style symbol” added above the musical score in FIG. 2A exemplarily shows the playing style of the music phrase shown in the musical score. Such a musical score with a “playing style symbol” is not indispensable at the time of sampling in this step S1. A musician plays the music phrase in accordance with a normal music score, and then analyzes the sampled waveform data to determine a performance style in each performance phase according to the passage of time, and creates a music score with such performance style symbols. You may do so. As will be described later, the score with such a rendition style symbol is not useful for sampling in step S1, but rather, a general user can obtain a desired score from a database created based on the data sampled here. It is thought that it will greatly assist the general user in extracting data and connecting them to create a desired performance sound. However, FIG.
In order to exemplify how the phrase shown in the musical score is played, the meaning of the rendition style symbols illustrated in the figure will be described here.

The performance symbol of a black circle drawn corresponding to the three notes in the first bar indicates the "star cart" performance style.
The size of the black circle indicates the volume level. The playing style symbol drawn with the letters "Atack-Mid, No-Vib" corresponding to the next note describes the playing style "medium attack, no vibrato". "Atk-Fast, Vib-Soon-Fast,
Release-Smoothly ''
"The attack rises quickly, the vibrato goes fast, and the release goes smoothly." A playing style symbol consisting of an oval black circle in the third measure indicates a “tenuto” playing style. In the third measure, a performance style symbol indicating that the volume is gradually reduced and a performance style symbol indicating that vibrato is added to the end of the sound are also described. As described above, it can be understood that various playing styles or performance expressions, that is, articulations are used even for a musical phrase having a length of about three measures. In addition,
The way of representing these rendition style symbols is not limited to this, but it is sufficient that the rendition style can be expressed in some way. Symbols expressing a certain degree of playing style are also used in conventional music notation, but in practicing the present invention, it is desirable to employ more precise performance style symbols than ever before.

In FIG. 1, the next step S2 is a step of dividing a series of sampled performance sounds into a plurality of time sections each having a variable length in accordance with the characteristic (ie, articulation) of the performance expression. It is. This is completely different from the method of dividing and analyzing the waveform data at regular regular time frames as known by Fourier analysis, for example. In other words, since the articulations present in a series of sampled performance sounds are diverse, the time range of the sound corresponding to each articulation is not a uniform time length but an arbitrary variable. Of length.
Therefore, dividing a series of sampled performance sounds into a plurality of time sections in accordance with the characteristics (ie, articulation) of the performance expression means that the length of each of the divided time sections is variable. Becomes

FIGS. 2 (b), 2 (c) and 2 (d) show examples of such a division of a time section in a hierarchical manner.
FIG. 2B shows a relatively large lump of articulation (referred to as "articulation large unit" for convenience, AL # 1, AL # 2, AL # 3, AL #
(Indicated by four symbols). Such a large articulation unit may be divided into, for example, small phrasing units having a common rough performance expression. FIG. 2C shows an example in which one large unit of articulation (AL # 3 in the figure) is further divided into units during articulation (for convenience, indicated by symbols AM # 1 and AM # 2). ing. The articulation units AM # 1 and AM # 2 are roughly divided into, for example, one sound. FIG.
One articulation unit (AM # 1, in the figure)
AM # 2) is further divided into minimum articulation units (for convenience, indicated by symbols AS # 1 to AS # 8). The minimum unit of articulation AS # 1 to AS # 8 is a part of a sound which is different in performance expression, typically an attack part, a body part (a relatively stable part showing a steady characteristic of a sound). ), The release section, and the connection between sounds.

For example, AS # 1, AS # 2, and AS # 3 form an attack portion, a first body portion, and a second portion of one sound (preceding sound of a slur) constituting unit AM # 1 during articulation. AS # 5, AS corresponding to each body part
# 6, AS # 7, and AS # 8 constitute one unit sound AM # 2 during the next articulation (sound following the slur). First body part, second body part, and third body part. , And the release section respectively. First and second
The reason why there are multiple body parts, such as the body part, is that even if the body part has the same sound, the articulation may be different (for example, the speed of vibrato etc. may change). It corresponds to such a case.
AS # 4 corresponds to a part of connection between sounds due to a slur change. This part AS # 4 is either one of the two articulation units AM # 1 and AM # 2 (the end part of AM # 1 or A
M # 2). Alternatively, the part AS # 4 of the connection between sounds due to such a slur change may be taken out as a unit during articulation from the beginning. In this case, the large articulation unit AL # 3 is divided into three articulation units, and the middle articulation unit, that is, the part of the connection between sounds, is directly articulated. Minimum unit AS
This corresponds to # 4. When the part AS # 4 of the connection between the sounds due to the slur change is taken out alone, the part AS # 4 is also used as the part connecting the other sounds with the sound, so that these parts AS # 4 are also used. Sounds can be connected by slurs.

The minimum articulation unit AS # 1 to AS # 8 as shown in FIG.
Corresponds to a plurality of time sections divided by the processing of. In the following, such an articulation minimum unit is also referred to as an articulation element. Note that the way of dividing the minimum unit of articulation is not limited to the above example, so the minimum unit of articulation, that is, the articulation element does not always correspond to only the sound part.

In FIG. 1, the next step S3 is a step for each divided time section (minimum articulation unit AS).
In this step, waveform data for each of # 1 to AS # 8, that is, articulation elements) is analyzed for a plurality of predetermined tone elements, and data indicating characteristics of the analyzed tone elements is generated. The musical tone elements to be analyzed include, for example, waveform (tone), amplitude (volume), pitch (pitch), time, and the like. These musical sound elements are components (elements) of the waveform data in the time section and also articulation elements (elements) in the time section. Next step S4
Then, data indicating the characteristics of each generated element is stored in a database. The database uses these accumulated data as template data when synthesizing music.
Make it available. An example of how to analyze these tone elements is as follows, and FIG. 3 shows an example of data (template data) indicating the characteristics of each tone element. FIG. 2E also shows examples of the types of each tone element analyzed from one minimum unit of articulation.

For the waveform (tone) element, the original PCM waveform data in the time section (articulation element) is taken out as it is. This is stored in the database as a waveform template (Timbre template). “Timbre” will be used as a symbol indicating this waveform (tone color) element. As for the amplitude (volume) element, the volume envelope (volume amplitude change over time) of the original PCM waveform data in the time section (articulation element) is extracted to obtain amplitude envelope data. This is stored in a database as an amplitude template (Amp template). “Amp” (abbreviation of Amplitude) is used as a symbol indicating the amplitude (volume) element. As for the pitch (pitch) element, the pitch envelope (pitch change with time) of the original PCM waveform data in the time section (articulation element) is extracted to obtain pitch envelope data. This is stored in a database as a pitch template (Pitch template). “Pitch” is used as a symbol indicating the pitch element.

For the time element, the time length of the original PCM waveform data in the time section (articulation element) is used as it is. Therefore, the original time length of the section (variable value)
Is expressed by the ratio “1”, it is not necessary to analyze and measure this time length when creating the database.
In this case, since the data on the time element, that is, the time template (TSC template) has the same value “1” in any section (articulation element), it is not necessary to store this in the template database. Of course, the present invention is not limited to this, and a modified example in which the actual time length is analyzed and measured, and this is stored in the database as time template data can be implemented.

As a technique for variably controlling the original time length of the waveform data, control for expanding or compressing the waveform data in the time axis direction without affecting the pitch of the waveform data is not disclosed. , "Time Str
It has already been proposed by the present inventors as "etch &Compress" control (abbreviated as "TSC control"). In the present embodiment, such “TSC control” is used, and the TSC used as a symbol of the time element is this abbreviation. At the time of musical tone synthesis, this TSC value is not fixed to “1”, but is set to any other appropriate value.
The time length of the reproduced waveform signal can be variably controlled. In this case, the TSC value may be given as a time-varying value (for example, an appropriate time function such as an envelope). Note that this TSC control can be useful when the time length of a portion of the original waveform to which a special playing style such as vibrato or slur is applied is variably controlled.

The above-described processing is performed for various natural musical instruments in various playing styles (for various musical phrases), and for each natural musical instrument, a large number of articulation elements are used for each musical tone element. Create templates and store them in a database.
Not only natural instruments, but also various sounds that exist in the natural world, such as human voices and thunder, perform sampling and articulation analysis as described above. Template data may be stored in a database. Of course, the phrase to be played live for sampling is not limited to a phrase consisting of several measures as in the above example, but may be shorter if necessary (for example, one phrasing small unit as shown in FIG. 2B). ) Alone or, conversely, an entire song.

The structure of the database DB is roughly divided into a template database TDB and an articulation database ADB, for example, as shown in FIG.
As the hardware of the database DB, a readable / writable storage medium (preferably a large-capacity medium) such as a hard disk device or a magneto-optical disk device is used as is well known. The template database TDB stores a large number of template data created as described above. Note that the template database TD
All of the template data stored in B need not necessarily be based on the sampling and analysis of performance sounds or natural sounds as described above. In short, the template data may be prepared in advance as a template (finished data). It may be created arbitrarily and artificially by data editing work. For example, the TSC template for the time element is normally “1” as described above as long as it is based on the sampled performance sound, but can be created with a free change pattern (envelope). From this, various TSC values or envelope waveforms of temporal changes thereof may be created as TSC template data and stored in a database. Further, the types of templates stored in the template database TDB are not limited to those corresponding to the specific elements analyzed from the original waveform as described above.
Other types may be increased as appropriate for convenience in tone synthesis. For example, when tone color control is performed using a filter at the time of musical tone synthesis, a large number of filter coefficient sets (including time-varying filter coefficient sets) are prepared as template data, and these are stored in the template database T.
It may be stored in a DB. Of course, such a filter coefficient set may be created based on the analysis of the original waveform, or may be created by other appropriate means.

The data structure of each template data stored in the template database TDB is composed of data representing the content of each template data as exemplified in FIG. For example, the waveform (Timbre) template is the PCM waveform data itself. In addition, amplitude (Amp) envelope, pitch (Pitch) envelope,
An envelope waveform such as a TSC envelope may be obtained by PCM-encoding the envelope shape.
However, in order to compress the data storage configuration of the template of the envelope waveform shape in the template database TDB, parameter data for approximating the envelope waveform with a polygonal line (data indicating the gradient rate of each polygonal line and a target level or time, etc., as is well known) These template data may be stored in the form of

The waveform (Timbre) template is also P
The data may be stored in an appropriate data compressed format other than the CM waveform data. Further, the waveform, that is, the timbre (Timbre) template data may be stored in another appropriate data format. That is, the waveform (Timbr
e) The template data is, for example, DPCM or ADP
The data may be waveform data in a coded format other than the PCM format such as CM or the like, or may be waveform formation data that does not directly indicate a waveform sample value, that is, waveform synthesis parameters. May be. As a waveform synthesis method using such parameters,
Various methods such as Fourier synthesis, FM (frequency modulation) synthesis, AM (amplitude modulation) synthesis, physical model sound source or SMS waveform synthesis (waveform synthesis using a deterministic component and an uncertain component) are known. May be adopted, and the parameters for waveform synthesis for that purpose may be stored in the database as waveform (Timbre) template data. In that case,
The waveform forming process based on the waveform (Timbre) template data, that is, the waveform synthesizing parameter, is of course performed by the corresponding waveform synthesizing arithmetic unit or program. In that case, a plurality of parameter sets for waveform synthesis for forming a waveform of a desired shape are stored in correspondence with one articulation element, that is, a time section, and a parameter set used for waveform synthesis is stored in time. By changing over as time passes, time variation of the waveform shape within one articulation element may be realized.

Further, the waveform (Timbre) template is defined as P
Even when the data is stored as the CM waveform data, if a known loop reading technique can be adopted (for example, waveform data of a portion such as a body portion in which the timbre waveform is stable and does not change much over time), Instead of storing the entire waveform of the section, only a part of the waveform data may be stored. If the contents of template data for different time intervals obtained as a result of sampling and analysis, that is, for the articulation elements, are the same or similar, only one template data is stored in the database TDB without being stored. Can be stored and shared at the time of musical tone synthesis, so that the storage amount of the database TDB can be saved. The configuration of the template database TDB may include a preset area created in advance by a supplier of the basic database (for example, an electronic musical instrument maker), a user area that can be freely created by a user, and the like.

Articulation Database AD
B constructs a performance containing one or more articulations, the data describing the articulation (ie the data describing a series of performances by a combination of one or more articulation elements and each articulation). (Data describing the translation elements) are stored in correspondence with various performance cases and playing styles. The block in FIG. 4 illustrates a database configuration for a certain instrument sound named “Instrument 1”. The articulation element sequence AESEQ describes a performance phrase including one or more articulations (ie, an articulation performance phrase) in the form of sequence data that sequentially indicates one or more articulation elements. Is what you do. For example, the articulation element sequence corresponds to the time-series order of the minimum unit of articulation (articulation element) as shown in FIG. 2D analyzed in the sampling and analysis process. It is.
A large number of articulation element sequences AESEQ are stored so as to cover various possible playing styles when playing the instrument sound. In addition, one articulation element sequence AESE
Q is “a small unit of phrasing” (a large unit of articulation AL # 1 and AL #) as shown in FIG.
2, AL # 3, AL # 4), or these “phrasing small units” (AL # 1,
AL # 2, AL # 3, AL # 4), or “articulation unit” (AM # 1, AM # 2) as shown in FIG. 2 (c).
Or may correspond to some of these “articulation units” (AM # 1, AM # 2).

The articulation element vector AEVQ is the instrument sound (Instrument)
For 1), the index of template data for each musical tone element for all articulation elements prepared (stored) in the template database TDB is provided in the form of vector data designating each template (for example, the template database T
(In the form of address data for extracting a required template from the DB). For example,
As shown in the examples of FIGS. 2D and 2E, corresponding to a certain articulation element AS # 1, each element (waveform, amplitude) constituting a partial musical tone corresponding to the articulation element AS # 1 , Pitch, and time), respectively, which stores vector data (this is referred to as an element vector) that specifically designates four templates Timbre, Amp, Pitch, and TSC.

In one articulation element sequence (performance style sequence) AESEQ,
Indexes of a plurality of articulation elements are described in the order of performance, and a set of templates constituting each articulation element described therein can be derived by referring to the articulation element vector AEVQ. It has become. FIG. 5 (a) shows several articulation element sequences AESE
An example of Q # 1 to AESEQ # 7 is shown. Explaining how to read this figure, for example, AESEQ # 1 =
(ATT-Nor, BOD-Vib-nor, BOD-
Vib-dep1, BOD-Vib-dep2, REL
-Nor) is the sequence AESE of sequence number 1
Q # 1 is ATT-Nor, BOD-Vib-nor,
BOD-Vib-dep1, BOD-Vib-dep
2, REL-Nor, which indicates a sequence of five articulation elements. The meaning of the index symbol of each articulation element is as follows.

ATT-Nor is "Normal Attack"
(The playing style in which the attack part stands up as a standard). BO
D-Vib-nor is "Body Normal Vibrato"
(A playing technique in which a standard vibrato is attached to the body). BOD-Vib-dep1 indicates "Body Vibrato Depth 1" (a playing style in which a vibrato that is one step deeper than the standard is attached to the body part). BOD-Vib-d
ep2 is "Body Vibrato Depth 2" (playing technique in which the body part is given a vibrato two levels deeper than the standard)
Is shown. REL-Nor indicates "normal release" (a playing style in which the release part falls down as standard).

Therefore, the sequence AESEQ # 1 starts with a normal attack, in which a normal vibrato is applied first in the body portion, then the vibrato is slightly deeper, then the vibrato is further deepened, and finally, a standard vibrato is applied in the release portion. It consists of an articulation that shows a falling sound. The articulation of the other sequences AESEQ # 2 to AESEQ # 6 shown by way of example can be similarly understood from the symbolic representation of the articulation element in FIG. 5A. Figure 5 for reference
The meaning of the symbols of some other articulation elements shown in FIG.

BOD-Vib-spd1 indicates "body vibrato speed 1" (a playing style in which a vibrato one step faster than the standard is attached to the body). BOD-Vi
b-spd2 indicates "Body Vibrato Speed 2" (a playing style in which a vibrato two times faster than the standard is applied to the body part). BOD-Vib-d & s1 is "Body-
Vibrato Depth & Speed 1 "(the playing style in which the depth and speed of the vibrato attached to the body part are each raised by one step from the standard). BOD-Vib-bri is "Body-
Vibrato Brilliant "(playing style with vibrato on the body and flashing its tone). BOD-
Vib-mld1 is "Body Vibrato Mild 1"
(A vibrato is attached to the body and the playing style is slightly milder). BOD-Cre-nor indicates "body normal crescendo" (playing style in which a standard crescendo is attached to the body). BOD-C
re-vol1 indicates "body crescendo volume 1" (a playing style in which the volume of the crescendo attached to the body portion is increased by one level). ATT-Bup-nor indicates "attack / bend-up normal" (playing style in which the pitch of the attack portion is bend-up at a standard depth and speed). REL-Bdw-nor indicates "Release Bend Down Normal" (playing method in which the pitch of the release part is bent down at a standard depth and speed).

Accordingly, the sequence AESEQ # 2 starts with a normal attack, in which a normal vibrato is applied first in the body portion, then the vibrato speed is slightly increased, then the vibrato speed is further increased, and finally in the release portion. It responds to articulations (playing techniques) that show the change of showing a standard falling sound. Also, the sequence AESEQ # 3
Corresponds to an articulation (playing style) that shows a change that gradually increases the depth of the vibrato and gradually increases the speed. Sequence A
ESEQ # 4 corresponds to articulation (playing style) that changes the sound quality (tone color) of the waveform at the time of vibrato. Sequence AESEQ # 5 corresponds to articulation (playing style) with crescendo. The sequence AESEQ # 6 corresponds to an articulation (playing style) in which the pitch of the attack portion rises (the pitch gradually increases). Sequence A
ESEQ # 7 corresponds to an articulation (playing style) in which the pitch of the release section is bed-down (the pitch gradually decreases). The articulation element sequence (reproduction style sequence) is not limited to the above, and may have many more types.

FIG. 5B shows an example of the configuration of an articulation element vector AEVQ for some articulation elements. Explaining how to read this diagram, vector data indicating a template corresponding to each element is described in parentheses. The leading symbol in each vector data indicates the type of the template. That is,
Timb indicates a waveform (Timbre) template, Amp indicates an amplitude (Amp) template, Pit indicates a pitch (Pitch) template, and TSC indicates a time (TSC) template. Show.

For example, ATT-Nor = (Timb-A
-Nor, Amp-A-nor, Pit-A-nor,
TSC-A-nor) is an articulation element ATT-No having a meaning of "normal attack".
r is Timb-A-nor (standard waveform template for attack part), Amp-A-nor (standard amplitude template for attack part), Pit-A-nor (standard pitch template for attack part) , TSC-A
-Nor (standard TSC template for attack part)
Are synthesized by the four templates.

As another example, the articulation element BOD-Vib-dep1 having the meaning of "body vibrato depth 1" is Timb-B-vi.
b (waveform template for body vibrato), A
mp-B-dp3 (amplitude template for vibrato depth 3 of body part), Pit-B-dp3 (pitch template for vibrato depth 3 of body part), TSC-B
The waveform is synthesized by four templates -vib (TSC template for vibrato of body part). As another example, the articulation element REL-Bdw-nor having the meaning of "release bend down normal" is Timb-R-bdw.
(Waveform template for release section bend down),
Amp-R-bdw (amplitude template for bend down of release section), Pit-R-bdw (pitch template for bend down of release section), TSC-R-
Waveforms are synthesized by four templates bdw (TSC template for bend down of the release section).

In order to facilitate the editing of the articulation, the attribute information ATR schematically describing the features of each articulation element sequence.
To each articulation element sequence A
It is good to memorize it attached to ESEQ. Similarly, it is preferable to store attribute information ATR which roughly describes the characteristics of each articulation element, attached to each articulation element vector AEVQ. In short, such attribute information ATR is stored in each articulation element (FIG. 2).
(D) is a description of the feature of (the minimum articulation unit as shown in (d)). An example of an articulation element related to an attack part is shown as an example of a symbol (index) of the articulation element, contents of each attribute information ATR, and each vector data indicating a template of each musical tone element. 6 is shown.

In the example of FIG. 6, the attribute information ATR is also managed in a hierarchical manner. That is, articulation elements related to the attack portion are all provided with common attribute information of “attack”, of which standard elements are further provided with attribute information of “normal”. The element to which the bend-up performance is applied is provided with attribute information "bend-up", and the element to which the bend-down performance is applied is provided with attribute information "bend-down". Furthermore, among the elements to which the bend-up playing technique is applied, "normal"
Attribute information is given, and for those having a bend depth shallower than the standard, attribute information of "depth / shallow" is given. For those having a bend depth greater than the standard, "depth / deep" Attribute information is added, and if the bend speed is slower than the standard
Attribute information such as "slow" is given, and attribute information "speed / fast" is given to items having a bend speed higher than the standard. Although illustration is omitted, similarly subdivided attribute information is added to the element to which the bend down playing style is applied.

FIG. 6 also shows that some template data is shared between different articulation elements. In FIG. 6, the vector data (in other words, the template index) of the four templates described in the column of each index (articulation element index) of the playing style is used to form a partial sound of the articulation element. 5 shows the vector data indicating the template, and the reading method is the same as in FIG. 5B. Here, in the element having the bend-up attribute, an element with an = sign means that the same template as that in the normal state is used. For example, the waveform (Timbre) template for the bend-up playing technique all uses the same waveform template Timb-A-bup for the bend-up normal.
The amplitude (Amp) template for the bend-up playing technique is all the same as the amplitude template Amp-A-bup for the bend-up normal. This is because, even if the bend-up playing technique is slightly changed, it is possible to use a common waveform without changing the waveform and amplitude envelope, and there is no problem in sound quality. On the other hand, the pitch (Pi
tch) The template must be different depending on the depth of the bend-up playing technique. For example, in the articulation element ATT-Bup-dp1 having the attribute of “depth / shallow”, in order to indicate a pitch template (a pitch envelope template corresponding to a shallow bend-up characteristic) corresponding thereto, Vector data Pit-A-dp1 indicating a pitch envelope template corresponding to shallow bend-up characteristics is used.

By sharing the template data in this way, the storage amount of the template database TDB can be reduced. Also, it is not necessary to record live performances for all playing styles when creating the database. Referring to FIG. 6, it can be understood that the speed of the bend-up playing technique is adjusted by making the time (TSC) template different. Since the pitch bend speed corresponds to the time required to reach from the predetermined initial pitch to the target pitch, the original waveform data has a predetermined pitch bend characteristic (bend from the predetermined initial pitch to the target pitch within a certain time). If the time length of the original waveform data is variably controlled by the TSC control, the time required to reach from the initial pitch to the target pitch, that is, the bend speed can be adjusted. Such variable control of the waveform time length based on the time (TSC) template is suitable for adjusting the speed of various playing styles such as the speed of tone rise, the speed of slurs, and the speed of vibrato. For example, a change in pitch in a slur can be realized by a pitch (Pitch) template, but a natural slur change can be realized by performing TSC control using a time (TSC) template.

Articulation database AD
It is assumed that the articulation element vector AEVQ in B can be addressed by the articulation element index, and can also be addressed by the attribute information ATR. Thus, by performing a search on the articulation database ADB using the desired attribute information ATR as a keyword, it is possible to search for what articulation element having an attribute corresponding to the keyword, It is convenient for data editing work by the user. Such attribute information ATR is preferably added to the articulation element sequence AESEQ. Thus, by searching the articulation database ADB using the desired attribute information ATR as a keyword, an articulation element sequence AESEQ including an articulation element having an attribute corresponding to the keyword can be searched. it can. The articulation element index for addressing the articulation element vector AEVQ in the articulation database ADB is given according to the reading of the articulation element sequence AESEQ. For work or real-time free sound creation, the desired articulation element index may be independently addressable.

Articulation database AD
In B, there is also an area for storing a user articulation element sequence URSEQ so that a user can create a desired articulation element sequence and store it. In such a user area, articulation element vector data created by the user may also be stored. The articulation database ADB stores a partial vector PVQ as vector data below the articulation element vector AEVQ. If the template data specified by the articulation element vector AEVQ is stored in the template database TDB not as data of the entire time section of the articulation element but as part of the data, The template data is read out in a loop (repeatedly read out) to reproduce the data of all the time sections of the articulation element. The data required for such a loop read is stored as a partial vector PVQ. In that case,
For example, the articulation element vector A
The EVQ stores data designating the partial vector PVQ in addition to the template data.
The data of the partial vector PVQ is read by the partial vector instruction data, and the loop reading is controlled by the data of the partial vector PVQ. Therefore, the partial vector PVQ includes data indicating a loop start address, a loop end address, and the like necessary for loop read control.

Further, in the articulation database ADB, rule data RULE describing rules for connecting waveform data between temporally adjacent articulation elements at the time of musical sound synthesis.
I remember. For example, to perform smooth cross-fading interpolation between waveform-adjacent articulation elements, connect directly without performing cross-fading interpolation, or perform cross-fading waveform interpolation Are stored for each sequence, or for each articulation element in the sequence. This connection rule can also be targeted for data editing by the user. In the articulation database ADB, an articulation database having a data structure as exemplified above is provided for each instrument sound (natural instrument tone), and various human voices (young female voice, young female voice, Male voice, baritone, soprano, etc.)
In addition, various kinds of natural sounds (thunder sounds, wave sounds, etc.) are provided.

[Outline of Musical Sound Synthesis] FIG. 7 shows an outline of the procedure for synthesizing musical sounds using the database DB created as described above. First, a required rendition style sequence corresponding to a musical tone performance to be generated (a performance phrase composed of a plurality of tones or one tone) is designated (step S1).
1). The instruction of the rendition style sequence is made of one of the articulation element sequence AESEQ or URSEQ of the desired musical instrument sound (or human voice sound or natural sound) stored in the articulation database ADB.
One may be selectively indicated.

The instruction of such a rendition style sequence (ie, articulation element sequence)
The information may be provided based on a real-time performance operation by a user, or may be provided based on automatic performance data. In the former case, for example, various rendition style sequences are assigned in advance to a keyboard or other performance operators, and the rendition style instruction data assigned thereto is generated in accordance with the operation of the operators. can do. In the latter case, as one method, as shown schematically in FIG.
Incorporating and storing rendition style instruction data as event data in automatic performance sequence data of the I format or the like, so that each rendition style sequence instruction data is read out at the time of predetermined event reproduction during automatic performance reproduction. Can be. In FIG. 8, DUR is duration data indicating a time interval until the next event,
EVENT indicates event data, MIDI indicates that the performance data attached to the event data is MIDI format data, and AESEQ indicates that the performance data attached to the event data is performance style sequence instruction data. In this case, an ensemble of an automatic performance based on the automatic performance data in the MIDI format or the like and an automatic performance based on the performance style sequence according to the present invention can be performed. In this case, for example, the main solo or melody musical instrument part may be played by a playing style sequence according to the present invention, that is, articulation element synthesis, and the other musical instrument parts may be played by automatic performance based on MIDI data. .

As another method of the latter, as schematically shown in FIG. 8B, only a plurality of performance style instruction data AESEQ are stored in an event data format corresponding to a desired music. This may be read at the time of reproduction of each predetermined event. As a result, it is possible to perform an articulation sequence automatic performance of a music, which has not been available in the past. Furthermore, as another method of the latter, only automatic performance sequence data of a MIDI format or the like corresponding to a desired music piece is stored, and the automatic performance sequence data is analyzed by a performance interpretation program, so that each phrase or note is analyzed. May be automatically analyzed, and the rendition style instruction data may be generated as a result of the analysis. Further, as another method of indicating the rendition style sequence, the user inputs desired one or more pieces of attribute information,
By searching the articulation database ADB using this as a keyword, one or more articulation element sequences AESE
Q may be automatically listed, and a desired sequence may be selected and designated from the list.

In FIG. 7, the selected articulation element sequence AESEQ or URSE
In Q, an articulation element (AE) index is read out according to a predetermined performance order (step S12). Then, the articulation element vector (AEV) corresponding to the read articulation element (AE) index is read.
Q) is read (step S13). Then, the read articulation element vector (AE
VQ) is read from the template database TDB (step S).
14).

Then, waveform data (partial sound) of one articulation element (AE) is synthesized according to each read template data (step S15). The method of synthesizing the waveform is basically such that the PCM waveform data corresponding to the waveform (Timbre) template data is stored in a pitch (PiPi) from the template database TDB.
tch) read speed and time (T
SC) Read with a time length according to the template, and amplify the amplitude envelope of the read PCM waveform data (Amp)
Control according to the template. In this embodiment, since the waveform (Timbre) template data stored in the template database TDB has the pitch, amplitude envelope and time length of the sampled original waveform as they are, the pitch (Pitch) template and the amplitude ( Amp) template and time (TSC) template, if they have not changed from the original waveform sampled,
The PCM waveform data corresponding to the waveform (Timbre) template data stored in the template database TDB is read out as it is as the waveform data for the articulation element. Pitch (Pitch) template, amplitude (Amp) template, time (TSC)
When any one of the templates is changed from that of the sampled original waveform, the reading speed of the waveform (Timbre) template data stored in the template database TDB is variably controlled (pitch) according to the change. The readout time length is variably controlled (when the template is changed), or the amplitude envelope for the readout waveform is variably controlled (when the amplitude template is changed). When the above-described partial vector PVQ is applied to the articulation element AE, necessary loop read control is also performed.

Next, processing for sequentially connecting the waveform data of each articulation element synthesized as described above is performed. As a result, a series of performances composed of a time series combination of a plurality of articulation elements is performed. A sound is generated (step S16). The connection process here is controlled according to rule data RULE stored in the articulation database ADB. For example, when the rule data RULE instructs direct connection, the waveform data of each articulation element synthesized in step S15 may be simply switched in sequence according to the order in which they are generated. When the rule data RULE indicates a predetermined cross-fade interpolation,
The waveform data at the end of the preceding articulation element and the waveform data at the beginning of the following articulation element are subjected to cross-fade interpolation synthesis so that the waveforms are smoothly connected. For example, if the sampled original waveform is connected as it is, it is guaranteed that each articulation element is connected smoothly from the beginning, so
The rule data RULE may indicate a direct connection.
Otherwise, there is no guarantee that articulation elements will connect smoothly,
Some kind of interpolation synthesis should be performed. As described later, any of a plurality of types of cross-fade interpolation formats can be arbitrarily selected by the rule data RULE.

A series of performance sound synthesizing processes as schematically shown in steps S11 to S16 are performed for one musical instrument sound (or human voice sound or natural sound) on one musical sound synthesis channel. If the performance sound synthesis processing for a plurality of instrument sounds (or human voice sounds or natural sounds) is to be performed simultaneously and in parallel, step S
A series of performance sound synthesizing processes as schematically shown in 11 to S16 may be performed on a plurality of channels in a time-division or parallel manner. As will be described later, when a musical tone waveform is formed using cross-fade synthesis processing, two waveform generation channels (a channel for generating a fade-out waveform and a waveform for fading in) are generated for one musical sound synthesis channel. Channel).

FIG. 9 schematically shows examples of combinations of articulation elements in some rendition style sequences. The rendition style sequence # 1 shown in (a) shows the simplest combination example, and the articulation element A in the attack portion
# 1, body part articulation element B
# 1, an articulation element R # 1 of a release section is sequentially connected, and a connection portion between the elements is cross-fade interpolated. A playing style sequence # 2 shown in (b) shows an example of an articulation combination in which a decoration sound is added before the main sound, and the articulation element A # 2 of the decoration sound attack portion and the decoration sound Articulation element B # 2 in the body part, articulation element A # 3 in the attack part for the main sound, articulation element B # 3 in the body part for the main sound, articulation in the release part for the main sound Curation elements R # 3 are sequentially connected,
The connection between the elements is cross-fade interpolated. A playing style sequence # 3 shown in (c) shows an example of an articulation combination in which the preceding sound and the following sound are connected by a slur. The articulation element A # 4 of the attack part for the preceding sound and the articulation element A # 4 for the preceding sound Articulation element B # 4 in body part, articulation element B # in body part for partial sound for slur
5, the articulation element B # 6 of the body portion for the subsequent sound and the articulation element R # 6 of the release portion for the subsequent sound are connected in order, and the connection between the elements is cross-fade. Interpolated. In the figure, the partial sound waveform corresponding to each articulation element is schematically shown only by an envelope for convenience, but actually, as described above, the waveform (Timbre), the amplitude (Amp), the pitch (Pitch) , And time (TSC) template data.

FIG. 10 is a time chart showing a specific example of processing for sequentially generating partial sound waves corresponding to a plurality of articulation elements and cross-fading them in one musical sound synthesis channel. Specifically, two waveform generation channels are used to cross-fade synthesize two element waveforms per tone synthesis channel. FIG. 10A shows an example of waveform generation in the first waveform generation channel, and FIG. 10B shows an example of waveform generation in the second waveform generation channel. In (a) and (b), the “synthesized waveform data” shown in the upper part of each of the waveforms (Timbre) and (Timbre) as described above as a partial sound waveform corresponding to the articulation element
Waveform data synthesized based on template data such as amplitude (Amp), pitch (Pitch), and time (TSC) (for example, waveform data synthesized in step S15 of FIG. 7) is shown at the bottom of each. The illustrated “cross-fade control waveform” indicates a control waveform used for cross-fading partial sound waveforms corresponding to each element. This “cross-fade control waveform” is formed in the process of step S16 in the flow of FIG. 7, for example. By controlling the amplitude of the upper element waveform data by the lower crossfade control waveform of each channel and adding the crossfade amplitude controlled waveform data of each channel (first and second waveform generation channels), Crossfade synthesis is completed.

When one rendition style sequence is started, a sequence start trigger SST is applied, and in response to this, the synthesis of the partial sound waveform corresponding to the first articulation element (provisionally A # 1) of the sequence is performed. Be started. That is, the waveform (Timbre), amplitude (Amp),
Waveform data is synthesized based on template data such as pitch (Pitch) and time (TSC). Therefore, in the figure, “synthesized waveform data” is simply indicated by a block, but actually, a waveform corresponding to the waveform (Timbre) template data, an amplitude envelope corresponding to the amplitude (Amp) template data, and Pitch
It has a pitch corresponding to the template data, its temporal change, and a time length corresponding to the time (TSC) template data. The rising of the crossfade control waveform may be such that the first articulation element waveform of the sequence immediately rises at the full level as shown. However, if it is desired to perform crossfade synthesis with the waveform at the end of the performance sound of the preceding sequence, the rise of the first crossfade control waveform of the sequence may have a fade-in characteristic with an appropriate slope. The gradient of the fade-in is set by the fade-in rate FIR # 1.

In response to the first articulation element A # 1 in the sequence, connection control information
It has the above-mentioned fade-in rate FIR # 1, next channel start point information NCSP # 1, fade-out start point information FOSP # 1, and fade-out rate FOR # 1. Next channel start point information NCSP # 1 indicates a point at which waveform generation of the next articulation element (for example, B # 1) is started. Fade-out start point information FOSP # 1 indicates a point at which fade-out of its own waveform is started. As shown in the figure, the crossfade control waveform indicates a full level flat up to the fadeout start point, but after the fadeout start point, the slope is in accordance with the set fadeout rate FOR # 1.
The level gradually falls. When the rule data RULE corresponding to the element A # 1 indicates direct connection without cross-fade connection,
These pieces of information NCSP # 1 and FOSP # 1 may indicate the end of the synthesized articulation element waveform. However, when the corresponding rule data RULE indicates direct connection without cross-fade connection, the information NCSP #
1, FOSP # 1 respectively designates appropriately set points before the end of the articulation element waveform, as shown in the figure. Therefore, the information NCSP # 1, FOSP # 1, FIR # 1, FOR
# 1 is the rule data R for the element A # 1
It may be considered to be included in ULE. The connection control information is provided for each articulation element.

The generation process of the element waveform A # 1 in the first waveform generation channel shown in FIG.
Next channel start point information NCSP # 1
When the point indicated by the arrow is reached, a next channel start trigger NCS # 1 is given to the second waveform generation channel shown in FIG. 10 (b), and the second articulation in the second waveform generation channel is performed. The generation of the partial sound waveform corresponding to the translation element B # 1 is started. Further, the articulation element B # 1
The cross-fade control waveform corresponding to (1) fades in (increases gradually) at the gradient set by the corresponding fade-in rate FIR # 2. Thus,
The fade-out period of the preceding articulation element A # 1 and the fade-in period of the following articulation element B # 1 overlap, and the cross-fade synthesis is completed by adding both.
After the waveform data of the preceding articulation element A # 1 fades out, only the succeeding articulation element B # 1 remains. Thus, the preceding articulation element A # 1
From the following articulation element B # 1
Is cross-faded to smoothly connect the waveforms.

The generation process of the element waveform B # 1 in the second waveform generation channel shown in FIG.
When the point indicated by the fade-out start point information FOSP # 2 is reached, as shown, the cross-fade control waveform changes to the set fade-out rate FO.
With the slope according to R # 2, the level gradually falls.
When the generation process of the element waveform B # 1 reaches a point indicated by the next channel start point information NCSP # 2, the next channel start trigger NCS # 2 generates the first waveform shown in FIG. Given to the channel and starts generating a partial sound waveform corresponding to the third articulation element R # 1 in the first waveform generation channel. Further, the crossfade control waveform corresponding to the articulation element R # 1 fades in (gradually rises) at the gradient set by the corresponding fade-in rate FIR # 3. In this way, the fade-out period of the preceding articulation element B # 1 and the fade-in period of the following articulation element R # 1 overlap, and the cross-fade synthesis is completed by adding both. Less than,
Similarly, each articulation element is connected in chronological order of the sequence while sequentially cross-fading.

In the above example, crossfade synthesis is performed on element waveforms synthesized based on each template. However, the present invention is not limited to this, and the crossfade processing may be performed for each template data, and the element waveforms may be synthesized based on the crossfade processed template data. In that case, different connection rules can be applied to each template even for the same element.
That is, each of the above connection control information (fade-in rate FIR, next channel start point NCS
P, fade-out start point FOSP, fade-out rate FOR) correspond to the waveform (Ti
mbre), amplitude (Amp), pitch (Pitch), time (TS
C) is prepared for each template corresponding to each tone element. In this way, crossfade connection can be performed for each template according to the optimum connection rule according to the template, which is effective.

[Editing] FIG. 11 schematically shows an example of the data editing process. In FIG. 11, a certain articulation element A # 1 having an attribute of an attack part, a certain articulation element B # 1 having an attribute of a body part, and a certain articulation having an attribute of a release part Articulation element sequence AES consisting of element R # 1
An example is shown in which editing is performed based on the data of EQ # x. Of course, in performing the data editing described here, a computer executes a required editing program,
The operation is performed by using an appropriate realizing means such that the user performs a desired operation using a keyboard or a mouse while checking the state of various data displayed on the display. The base sequence AESEQ # x can be selected from a number of sequences AESEQ (for example, see FIG. 5A) stored in the articulation database ADB. Editing of articulation data is roughly divided into replacement, addition or deletion of an articulation element in a sequence, and replacement of a template in an element or creation of a new template by modifying the data value of an existing template.

In the edit column of FIG. 11, the articulation element R # 1 of the release part in the base sequence AESEQ # x has a relatively gently falling amplitude envelope characteristic. An example in which an element R # x having a rapidly falling amplitude envelope characteristic is replaced is shown. Not only replacement, but also addition of a desired element (for example, addition of a body part element or addition of an element for a decorative sound) and deletion (if there are a plurality of body parts, any of them can be deleted) is also possible. is there. The element R # x used for replacement can be selected from a large number of articulation element vectors AEVQ (see, for example, FIG. 5B) stored in the articulation database ADB. In that case, attribute information A
A desired element R # x to be used for replacement can be selected from an element group having the same attribute with reference to TR.

Next, the template data corresponding to the desired tone element in the desired element (for example, the replaced element R # x) is replaced with another template data relating to the tone element. The example of FIG. 11 shows that the pitch (Pitch) template of the element R # x is replaced with another pitch template Pitch ′ (for example, a pitch template having a pitch bend characteristic).
As a result, the created new release element R # x ′ has an amplitude envelope characteristic that falls relatively quickly and a pitch bend-down characteristic. In the case of replacing a template, referring to the attribute information ATR, a template (vector data) of a group of elements having the same attribute in a large number of articulation element vectors AEVQ (for example, FIG. 5B) is selected. A desired template (vector data) to be used for replacement can be selected. The new element R # x ′ created by replacing some of the templates is given an articulation element vector AEVQ (see FIG. 4) of the articulation database ADB by adding a new index and required attribute information. It is advisable to additionally register in the area of).

It is also possible to modify the specific data content of a desired template. In this case, the specific data content of the desired template for the element being edited is read from the template database TDB, displayed on a display or the like, and the data content is appropriately changed by operating a keyboard or mouse. When the desired data correction is completed, the corrected template data is newly registered in the template database TDB with a new index, and new vector data is assigned to the corrected template data. A new index and required attribute information are added to a new element including data (for example, R # x ′) and additionally registered in the area of the articulation element vector AEVQ (see FIG. 4) of the articulation database ADB. It is better to do it.

As described above, the base sequence A
A data editing process for creating new sequence data by appropriately changing the contents of ESEQ # x can be performed. New sequence data created by such a data editing process is given a new sequence number (for example, URSEQ # x) and attribute information as a user articulation element sequence URSEQ, and registered in the articulation database ADB. Thereafter, when synthesizing a tone, the sequence number URSEQ #
Articulation database ADB using x
Can read data of the user articulation element sequence URSEQ. It should be noted that the form of data editing is not limited to the example illustrated in FIG. 11, and there may be various forms. For example, a desired element may be sequentially selected from the element vector AEVQ without calling the base sequence AESEQ, thereby creating the user sequence URSEQ.

FIG. 12 is a flowchart showing an outline of a computer program capable of executing the data editing processing as described above. In step S21, a desired performance style is specified. This designation may be made by directly inputting the sequence AESEQ or URSEQ number using a computer keyboard or mouse, or by inputting desired musical instrument timbre and attribute information.
In the next step S22, a sequence matching the specified playing style is stored in the articulation database ADB.
Search for AESEQ or URSEQ in the corresponding sequence AESEQ or URSE
Select Q. In this case, the sequence AESEQ or U
When the RSEQ number is directly input, the corresponding item is directly retrieved. If attribute information is input, the sequence AESEQ and / or URS corresponding to the attribute information
The EQ is searched. A plurality of pieces of attribute information can be input, and when a plurality of pieces of attribute information are input, for example, a search may be performed using AND logic. Of course, the search is not limited to this and may be performed using OR logic. The search results are displayed on a computer display, and a plurality of sequences AESEQ and / or URSEQ
Is searched, a desired one can be selected.

In step S23, the user is inquired whether or not to continue the editing operation. If NO (not to continue), the process goes to the exit and ends the editing process. Step S
If the content of the sequence selected or searched at 22 is as desired and there is no need for editing, the editing process ends. If it is desired to continue the editing process, step S23
Is YES, and the procedure goes to step S24. In addition, if a performance corresponding to the performance style specified in step S22 cannot be retrieved, it is determined that continuation is YES in step S23, and the process proceeds to step S24. Fig. 5 shows an example of retrieval using attribute information
An example in which data as shown in FIG. 6 is stored in the articulation database ADB will be described. For example, assume that “attack bend-up normal”, “body normal”, and “release normal” are input as attributes of the search condition of the articulation sequence. In this case, since the attribute matches the attribute of the sixth sequence AESEQ # 6 shown in FIG. 5A, the sequence AESEQ # 6 is searched and selected in step S22. If satisfied, the result is NO in step S23, and the editing process ends. If the user wants to continue the editing process, YES is determined in the step S23, and the process proceeds to the step S24.

In step S24, if a sequence corresponding to the performance style designated in step S21 has not been selected yet, a sequence closest to it is selected. For example, as the attributes of the search condition of the articulation sequence, "attack bend-up normal" and "vibrato normal" in step S21,
It is assumed that “Release Normal” is input. If there are only seven types of sequence AESEQ as shown in FIG. 5A, a sequence satisfying these conditions cannot be searched, and in step S24, the closest sequence AESEQ # 6 is selected. In step S25, the desired articulation element (A
A process of replacing vector data (index) indicating E) with vector data (index) indicating another articulation element is performed. For example, in the case of the above example, the element configuration of the sequence AESEQ # 6 selected as the closest sequence in step S24 is ATT-Nor, BOD-Nor, REL
-Nor (see FIG. 5A), so that the element B for the body part
OD-Nor (normal body) may be replaced with a vibrato body part element. For this purpose, referring to the articulation element vector AEVQ (for example, FIG. 5B), BOD-Vib-nor
Extract the element vector data (index) of (Body Normal Vibrato) and convert it to BOD-
Replace with Nor.

As needed, addition and deletion of articulation elements are also performed in step S25. When the replacement, addition, and / or deletion of the desired element vector data is completed, a new articulation element sequence is created (step S26). Since the replacement of the articulation elements and / or the addition or deletion of the articulation elements does not guarantee the connection of the waveforms between the elements in the newly created articulation element sequence, the connection rule data RULE is set in the next step S27. Set. In the next step S28, it is checked whether the set connection rule data RULE is sufficient. If it is not OK, the process returns to step S27, and the connection rule data RULE is set again. If the set connection rule data RULE is OK, the process proceeds to step S29.
In step S29, an inquiry is made as to whether to continue the editing process. If the editing process is not to be continued, step S
In step 30, the newly created articulation element sequence is registered in the articulation database ADB as a user sequence URSEQ. If the user wants to continue the editing process, Y
As ES, the process goes to step S24 or S31. In this case, if it is desired to return to replacement and / or addition / deletion of the articulation element, the process returns to step S24, and if it is desired to proceed to editing of template data, the process proceeds to step S31.

In step S31, the articulation element (A
Select E). In the next step S32, the template vector data corresponding to the desired tone element in the selected articulation element (AE) is replaced with another template vector data relating to the tone element. For example, as the attribute of the search condition of the articulation sequence, "attack bend up
"Normal", "Slightly slow vibrato", and "Release normal" are designated and input in step S21, and FIG.
It is assumed that AESEQ # 6 is selected in step S24 as the closest sequence among the sequences AESEQ shown in FIG. As described above, this sequence AESE
Since the element for the body part of Q # 6 is BOD-Nor (normal body), the element of the body part for vibrato, for example, BOD-Vi is used in step S25.
Replace with b-nor (body normal vibrato). Then, in step S31, this BOD-Vib-
Select an element of nor (body normal vibrato) and make it an object of editing. Then, in order to realize the desired “slightly slow vibrato”, step S 32
, Among the template vectors of BOD-Vib-nor (body normal vibrato), the time template vector TSC-B-vib is set to a time template vector (for example, TSC-B-sp2) for slightly lowering the vibrato speed. ).

In this way, among the templates of BOD-Vib-nor (body normal vibrato), the time template vector is converted from TSC-B-vib to TS
A new articulation element replaced with CB-sp2 is created (step S33). Also, a new articulation element sequence is created by replacing the body part element of the sequence AESEQ # 6 with the newly created articulation element (step S3).
3). Subsequent steps S34, S35 and S36 are the same as steps S27, S28 and S29 described above.
That is, the connection of the waveforms between the elements in the newly created articulation element sequence is not guaranteed by the replaced template data, so that the connection rule data RULE is set again as described above.

At step S36, an inquiry is made as to whether or not to continue the editing process. If the editing process is not to be continued, the process proceeds to step S37, and the newly created articulation element (AE) is registered in the articulation database ADB as a user articulation element vector (AEVQ).
If the editing process is to be continued, the determination in step S36 is YES, and the process proceeds to step S31 or S38. In this case, if it is desired to return to the replacement of the template vector, the process returns to step S31, and if it is desired to proceed to edit the specific contents of the template data, the process proceeds to step S38. In step S38, a template in a required articulation element (AE) whose data content is to be edited is selected. In a next step S39, the data of the selected template is stored in the template database TDB.
And the specific data contents are appropriately changed.

For example, "attack bend-up normal", "quite slow vibrato", and "release normal" are designated and input in step S21 as attributes of the search condition of the articulation sequence, and FIG. It is assumed that AESEQ # 6 is selected in step S24 as the closest sequence among the sequences AESEQ shown in FIG. As described above, this sequence AES
The element for the body part of EQ # 6 is BOD-Nor
(Normal body), so this is
In the body element for vibrato, for example BOD-
Replaced with Vib-nor (body normal vibrato). Then, in step S31, this BOD-Vi
An element of b-nor (body normal vibrato) is selected, and this is set as an editing target. Then, in order to realize the desired “substantially slow vibrato”, in step S32, among the template vectors of the BOD-Vib-nor (body normal vibrato),
A vector of a time template TSC-B-vib is converted to a vector of a time template (e.g., TSC-
B-sp1). However, in the time template indicated by this time template vector TSC-B-sp1, the desired "quite slow vibrato"
Is not realized yet, the time template vector TSC-B-sp1 is selected in a step S38, and the specific data content is changed to a content realizing a slower vibrato in a step 39. Also, new vector data (for example, TSC-B-sp0) is assigned to a new time template created by the change.

Thus, new time template data and its vector data TSC-B-sp0 are created (step S40). Further, a new articulation element (AE) in which the time template vector is changed to a new vector is created, and the element for the body part of the sequence AESEQ # 6 is replaced with the newly created articulation element (AE). A
A new articulation element sequence replaced with E) is created (step S4).
0). Subsequent steps S41, S42, S43 are the same as steps S27, S28, S29 described above.
That is, the connection of the waveforms between the elements in the newly created articulation element sequence is not guaranteed by the corrected template data, so that the connection rule data RULE is set again as described above.

In step S43, an inquiry is made as to whether or not to continue the editing process. If the editing process is not to be continued, the process proceeds to step S44, and the newly created template data is registered in the template database TDB. If the user wants to continue the editing process,
As S, the process returns to step S38. After step S44, the process proceeds to step S37, where the newly created articulation element (AE) is registered in the articulation database ADB as a user articulation element vector (AEVQ).
Further, the process proceeds to step S30, and the newly created articulation element sequence is registered as a user sequence URSEQ in the articulation database ADB. The procedure of the editing process is not limited to that shown in FIG.
Further, as described above, without calling the base sequence AESEQ, a desired element is sequentially selected from the element vector AEVQ, and the template data in each element is appropriately replaced or data is corrected, and based on this, the user The sequence URSEQ may be created. Although not specifically shown, it is preferable that a sound corresponding to the waveform of the articulation element being edited is generated at an appropriate stage of the editing process so that the user can confirm the sound by ear.

[Description of Partial Vector] FIG.
7 conceptually illustrates the concept of the partial vector PVQ. FIG. 13A schematically shows data (that is, normal template data) of an articulation element of a certain section, which is analyzed for a certain musical tone element (for example, a waveform). FIG. 13B shows partial template data PT1, PT1 distributedly extracted from the data of all sections shown in FIG.
2, PT3 and PT4 are schematically shown. These partial template data PT1, PT2, PT3, P
T4 is stored in the template database TDB as template data of the musical tone element. One template vector is assigned to this template data in the same manner as usual (similar to the case where data of all sections is stored as template data as it is). For example, if the template vector for this template data is “Timb-B-nor”, each partial data PT1, PT2, PT3, PT3
The template vector of No. 4 is “Timb-B-nor”
Is common. In this case, identification data indicating that the partial vector PVQ is included is registered as data attached to the template vector “Timb-B-nor”.

The partial vector PVQ indicates, for each of the partial template data PT1 to PT4, data indicating a storage position of the data in the template database TDB (for example, corresponding to a loop start address) and a width W of the data. It includes data (for example, corresponding to a loop end address) and data indicating a period LT in which the data is repeated. In the figure, for convenience, the width W and the period LT are illustrated as if they are common to all the partial data PT1 to PT4, but this is arbitrary for each of the data PT1 to PT4. Further, the partial template data PT1 to PT
The number of T4 is not limited to four and is arbitrary. Each partial template data PT based on the partial vector PVQ
By reading loops 1 to PT4 for the repetition period (LT), respectively, and connecting the read loops, it is possible to reproduce data of all sections as shown in (a). This reproduction process is called a decoding process. As an example of this decoding processing method,
The respective partial template data PT1 to PT4 may simply be read out in a loop for the repetition period LT.
Cross-fade synthesis may be performed while reading two waveforms in a loop. The latter is preferable because the connection between the loops is improved.

FIGS. 13 (c) and 13 (d) show examples of decoding processing by such cross-fade synthesis.
(C) shows an example of a crossfade control waveform in the first channel for crossfade synthesis, and (d) shows an example of a crossfade control waveform in the second channel for crossfade synthesis. That is, the first partial template data PT1 is faded out during the period LT with the fade-out control waveform CF11 shown in (c), and at the same time, the next partial template data PT2 is faded out with the fade-in control waveform CF21 shown in (d). Fade in during the period LT. Fade-out controlled data P
By adding T1 and the data PT2 subjected to the fade-in control, a loop read for cross-fading from the data PT1 to the data PT2 during the period LT is performed.
Next, the data PT1 is switched to the data PT3, the control waveform is switched to the fade-in waveform CF12, the control waveform of the data PT2 is switched to the fade-out waveform CF22, and cross-fade synthesis is performed. Thereafter, crossfade synthesis is performed by sequentially switching as shown in the figure. When performing crossfade synthesis, processing is performed so that the phases and pitches of the two loop readout waveforms match appropriately.

FIG. 14 is a flowchart showing an example of the template reading process in consideration of the partial vector PVQ. Steps S13 to S14c shown here correspond to the processing of steps S13 and S14 in FIG. In step S13, vector data of each template corresponding to the specified element is read from the data group of the articulation element vector AEVQ. In step S14a, it is checked whether or not the partial vector PVQ exists based on the identification data indicating that the partial vector PVQ exists. If there is no partial vector PVQ, the procedure goes to step S14b, where each template data is read from the template database TDB. If there is a partial vector PVQ, the process proceeds to step S14c, and the above-described “decoding process” is performed based on the partial vector PVQ. Thereby, the template data of the entire section for the element is reproduced (decoded).

When the partial vector PVQ is applied to a certain articulation element, it is not necessary to set the templates for all the tone elements of the articulation element as partial templates, and loop-read them as partial templates. It is sufficient to use a partial template only for a musical tone element of a type suitable for. In addition, the method of reproducing template data of the entire section of the element based on the partial vector PVQ is not limited to the simple loop reading as described above, and any other appropriate method may be used. For example, a partial template of a predetermined length corresponding to the partial vector PVQ is time-axis-extended as necessary, or a limited plurality of partial templates are combined at random or in a predetermined sequence, or the entire section of the element or A method of arranging over a required section may be used.

[Explanation of Vibrato Synthesis] Here, some new ideas on the method of vibrato synthesis will be described. FIG. 15 is a diagram schematically showing an example in which waveform data of a body part having a vibrato component is subjected to data compression by applying the concept of the partial vector PVQ, and a decoding example thereof. (A) illustrates an original waveform A including vibrato. In this original waveform, not only the waveform pitch fluctuates in one cycle of vibrato, but also the amplitude fluctuates. (B) shows a plurality of waveforms a1, dispersed from the original waveform of (a).
A state in which a2, a3, and a4 are taken out is exemplified. As these waveforms a1 to a4, those having different waveform shapes (tone colors) are selected, and one wavelength (one cycle of the waveform) is extracted as one or more waves with the same data size (number of addresses). These waveforms a1 to a4 are stored in the template database TDB as partial template data (that is, loop waveform data). This reading method is performed by sequentially reading out the waveforms a1 to a4 in a loop and performing crossfade synthesis.

FIG. 15C shows a pitch template whose pitch fluctuates during one vibrato cycle. Note that the pitch change pattern of this pitch template is a pattern which starts from a high pitch, shifts to a low pitch, and finally returns to a high pitch in the drawing, but is not limited to this, and other patterns (for example, from a low pitch to a high pitch) are used. A pattern that shifts and returns to a low pitch, or a pattern that starts from an intermediate pitch and returns to a high pitch → low pitch → intermediate pitch) may be used.

FIG. 15D shows each waveform a1 read out from the loop.
6 illustrates crossfade control waveforms for. At first, the waveforms a1 and a2 are loop-read (repeatedly read) at a pitch according to the pitch template of (c), and the amplitude control of fade-out is performed for the loop-read waveform a1 and fade-in is performed for the loop-read waveform a2. To combine the two. Thereby, the waveform a
From 1 to a2, the waveform shape cross-fades and changes sequentially, and the pitch of the cross-fade composite waveform changes sequentially at a pitch according to the pitch template. Thereafter, the waveforms are sequentially switched in the same manner, and crossfade synthesis is performed at a2 and a3, then at a3 and a4, and then at a4 and a1.

FIG. 15E shows the synthesized waveform data A ′.
Is shown. The waveform data A ′ is changed by smoothly cross-fading the waveform shape from the waveform a1 to a4 in order during one period of the vibrato, and the pitch changes according to the pitch template. Is attached. By repeating the process of synthesizing the waveform data A 'for one vibrato cycle as described above, waveform data over a plurality of vibrato cycles can be synthesized. In such a case, the pitch template for one cycle of vibrato as shown in (c) may be looped by the required number of vibrato cycles. Therefore, the structure of the partial vector PVQ may be hierarchical. That is, the hierarchical structure in which the waveforms a1 to a4 are individually read out in a loop as described above for synthesizing the waveform for one cycle of vibrato, and the whole (for one cycle of vibrato) is further repeated according to the looping of the pitch template. It may be. Note that the number of waveforms to be dispersedly extracted from the original waveform for vibrato synthesis is not limited to four as shown in FIG. 15B, and may be any number. In addition, the extraction intervals of each waveform need not be equal intervals. Further, the present invention is not limited to extracting from a range of one vibrato cycle as shown in FIG. 15B, and may be appropriately dispersedly extracted from a range of a plurality of vibrato cycles or a range of less than one vibrato cycle.

FIG. 16 is a diagram showing another example of another vibrato synthesis. In this example, a plurality of waveforms a1 to a4, b1 to b4, and c1 to c4 are distributed from sections A, B, and C over a plurality of vibrato periods of an original waveform including vibrato.
Take out c4. These waveforms a1 to a4, b1 to b
4, c1 to c4 select ones having different waveform shapes (tone colors) in the same manner as described above, and one wavelength (one cycle of the waveform).
Are extracted as one or more waves with the same data size (number of addresses). These waveforms a1 to a4, b1 to
b4, c1 to c4 are stored in the template database TDB as partial template data. This reading method is basically similar to the above-described example, in which each of the waveforms a1 to a
4, b1 to b4, c1 to c4 are sequentially read out in a loop, and cross-fade synthesis is performed. The difference from the above example is that the waveforms a1 to a4, b1 to b4 in the example of FIG.
4, c1 to c4 are interchanged with each other and the waveforms to be subjected to crossfade synthesis are arbitrarily combined so that various variations of the waveform timbre change in vibrato can be obtained in various combinations. is there.

For example, the waveforms a1 to a4, b1 to b4,
If the positions of these waveforms are exchanged without changing the relative time positions within one vibrato cycle of c1 to c4, for example, a1 → b2 → c3 → a4 → b1 → c2 → a
It is possible to obtain a replacement pattern of waveform positions such as 3 → b4 → c1 → a2 → b3 → c4. If vibrato synthesis processing by cross-fade synthesis similar to that of FIG. 15 is performed according to such a pattern of replacement of waveform positions, tone color changes different from vibrato obtained by vibrato synthesis processing by cross-fade synthesis according to the original waveform position pattern. You can get vibrato. The waveforms a1 to a4, b1 to b4, c
The reason why the positions of these waveforms are exchanged without changing the relative time positions within one vibrato cycle of 1 to c4 is to prevent unnaturalness due to the exchange. Such a replacement pattern of the waveform positions includes the twelve waveforms a1 to a4 and b1 shown in FIG.
B4, c1 to c4, 3 per vibrato cycle
There are 81 combinations of the fourth power, and there are 81 third power combinations in three cycles of vibrato. Therefore, the variation of the waveform timbre change in vibrato becomes extremely diverse. Which combination pattern is to be adopted may be selected at random.

A waveform having a vibrato characteristic created by the method shown in FIG. 15 or FIG. 16 (for example, FIG.
A ′) or a waveform having a vibrato characteristic created by another method, the vibrato characteristic can be variably controlled by a pitch (Pitch) template, an amplitude (Amp) template, and a time (TSC) template. it can. For example, the pitch template can control the depth of the vibrato, the amplitude template can control the depth of the amplitude modulation added with the vibrato, and the time (TS
C) Vibrato speed can be controlled (vibrato cycle can be controlled) by controlling the expansion and contraction of the time length of the waveform constituting one vibrato cycle using the template.

For example, in FIG. 15, the time length of each cross-fade section shown in (d) is subjected to time axis expansion / contraction control (TSC control) in accordance with a desired time (TSC) template, so that a musical sound reproduction pitch (waveform readout) is obtained. When the TSC control is performed without changing the address change rate, the time length of one cycle of vibrato can be controlled to expand and contract, whereby the vibrato frequency can be controlled.
In this case, if a TSC template is prepared corresponding to one vibrato cycle in the same manner as a pitch template as shown in (c), then the TSC template for one vibrato cycle is required for the required number of vibrato cycles. Just loop it. In addition, in conjunction with the time axis expansion / contraction control of the waveform according to the TSC template, the pitch (Pitch)
If the template and the amplitude (Amp) template are also subjected to time axis expansion and contraction control, the time axis expansion and contraction control can be performed in conjunction with these tone elements. By shifting the pitch change envelope characteristic indicated by the pitch template up and down, the tone reproduction pitch of the vibrato waveform can be variably controlled. In this case, by not performing the time axis control of the waveform using the TSC template, it is possible to control the time length of one vibrato cycle to be constant regardless of the musical tone reproduction pitch.

[Description of Connection Rule RULE] Next, a specific example of rule data RULE that describes how to connect articulation elements will be described.
For each tone element, for example, there are the following connection rules. (1) Waveform (Timbre) template connection rule Rule 1: Direct connection. Preset playing style sequence (articulation element sequence AES
In the case where smooth connection between the articulation elements is guaranteed in advance as in EQ), there is no problem in directly connecting without performing interpolation. Rule 2: interpolation in which the end of waveform A of the preceding element is extended. This interpolation example has a form as shown in FIG. 17A, in which the connection waveform C1 is synthesized by extending the end portion of the waveform A of the preceding element. The waveform B of the succeeding element is used as it is, and the connection waveform C1 extending to the end of the waveform A of the preceding element is faded out, and the beginning of the waveform B of the succeeding element is faded in to perform cross-fade synthesis. The connection waveform C1 is formed by repeating a one-period waveform or a plural-period waveform at the end portion of the waveform A of the preceding element by a required length.

Rule 3: interpolation in which the leading end of waveform B of the succeeding element is extended. This interpolation example has a form as shown in FIG. 17B, in which the leading end of the waveform B of the succeeding element is extended to synthesize the connection waveform C2. The waveform A of the preceding element is used as it is, and the end portion of the waveform A of the preceding element is faded out, and the connection waveform C2 is faded in to perform cross-fade synthesis. Also in this case, the connection waveform C2 is formed by repeating the one-period waveform or the plural-period waveform at the leading end of the waveform B of the succeeding element by a required length. Rule 4: Interpolation in which both the end portion of waveform A of the preceding element and the tip portion of waveform B of the succeeding element are extended. This interpolation example has a form as shown in FIG. 17C, in which the connection waveform C1 obtained by extending the end portion of the waveform A of the preceding element and synthesized and the connection waveform obtained by extending the front end portion of the waveform B of the subsequent element are synthesized. Cross-fade synthesis with the use waveform C2. In the case of Rule 4,
Since the time of the entire synthesized waveform is extended by the cross-fade synthesis period of C1 and C2, the time axis compression process is performed by the TSC control.

Rule 5: As shown in FIG. 17D, a connection waveform C prepared in advance is inserted between the waveform A of the preceding element and the waveform B of the succeeding element. At this time, the end portion of the waveform A of the preceding element and the tip portion of the waveform B of the succeeding element are partially removed by the connection waveform C. Alternatively, without deleting the end portion of the waveform A of the preceding element and the end portion of the waveform B of the succeeding element,
The connecting waveform C may be inserted, but in that case, the time of the entire synthesized waveform is extended, so that the time axis compression processing is performed by TSC control accordingly. Rule 6: As shown in FIG. 17 (e), a connection waveform C prepared in advance is inserted between the waveform A of the preceding element and the waveform B of the succeeding element, and at this time, the end of the waveform A of the preceding element The part and the first half of the connection waveform C are subjected to crossfade loss-fade synthesis, and the leading end of the waveform B of the succeeding element and the second half of the connection waveform C are subjected to crossfade loss-fade synthesis. Also in this case, if the time of the entire synthesized waveform is extended or shortened, the time axis compression processing is performed by the TSC control.

(2) Connection Rules for Other Templates Since data of templates (amplitude, pitch, and time) other than the waveform (Timbre) template take a simple form of an envelope waveform, two-channel cross-fade control is performed. Smooth connection can be realized by simpler interpolation processing without using complicated interpolation processing using waveforms. In particular, when interpolating and synthesizing the template data of the envelope waveform, the interpolation result is represented by a difference value (with a plus or minus sign) from the original template data value.
It is preferable to generate it. Then, an interpolation operation for smooth connection can be achieved only by adding a difference value (with a plus or minus sign) as an interpolation result to the original template data value read from the template database TDB in real time. It can be done and it is very simple. Rule 1: Direct connection. This example is shown in FIG. 1
Template for the th element (envelope waveform)
The end of AE1 and the template of the second element (envelope waveform) AE2-a have the same level, and the template of the second element (envelope waveform) AE2-a and the template of the third element (envelope waveform) Since the head level of the envelope waveform AE3 also matches, there is no need for interpolation.

Rule 2: Interpolation processing for smoothing in a local range before and after the connection point is performed. This example is shown in FIG.
Shown in AE1 to AE2-b in a predetermined range CFT1 between the end portion of the template (envelope waveform) AE1 of the first element and the tip portion of the template (envelope waveform) AE2-b of the second element.
Interpolation processing is performed so as to smoothly transition to. Also, the template (envelope waveform) A of the second element
Interpolation processing is performed so as to smoothly transition from AE2-b to AE3 within a predetermined range CFT2 at the end of E2-b and the template (envelope waveform) AE3 of the third element. The data E1 ', E2', E3 'obtained as a result of the interpolation are the original template values (envelope values) E1, E2, E
It is assumed that it is made up of a difference value (with a sign) for 3. By doing so, as described above, it is only necessary to add the difference values E1 ', E2', E3 'as interpolation results to the original template data values E1, E2, E3 read out from the template database TDB in real time. Therefore, an interpolation operation for smooth connection can be achieved, which is extremely simple.

A specific example of the interpolation processing of the rule 2 is shown in FIG.
As shown in FIGS. 9A, 9B, and 9C, there are a plurality of variations. In the example of FIG. 19A, the template data value EP at the end point of the preceding element AEn and the template data value SP at the start point of the succeeding element AEn + 1
In the interpolation area RCFT at the end of the preceding element AEn, the template data value of the preceding element AEn is set to the target value MP.
Interpolation is performed so as to approach. As a result, the locus of the template data of the preceding element AEn changes from the original line E1 to E1 '. In the interpolation area FCFT at the leading end of the succeeding element AEn + 1,
The template data value of the succeeding element AEn + 1 is started from the intermediate value MP, and interpolation is performed so as to gradually approach the original template data value locus indicated by the line E2. As a result, the locus of the template data value of the succeeding element AEn + 1 in the interpolation area FCFT is represented by the line E
As shown in 2 ′, the trajectory approaches the original trajectory E2.

In the example of FIG. 19B, the succeeding element A
Using the template data value SP at the start point of En + 1 as the target value, the interpolation area RC at the end of the preceding element AEn
In FT, interpolation is performed so that the template data value of the preceding element AEn approaches the target value SP. As a result, the locus of the template data of the preceding element AEn changes from the original line E1 to E1 ''. In this case, there is no interpolation area FCFT at the leading end of the succeeding element AEn + 1. In the example of FIG. 19C, the interpolation area FC at the leading end of the succeeding element AEn + 1 is used.
In the FT, the template data value of the succeeding element AEn + 1 is started from the template data value EP at the end point of the preceding element AEn, and interpolation is performed so as to approach the original locus of the template data value indicated by the line E2. As a result, the locus of the template data value of the succeeding element AEn + 1 in the interpolation area FCFT gradually approaches the original locus E2 as shown by the line E2 ''. In this case, there is no interpolation area RCFT at the rear end of the preceding element AEn. In FIG. 19 as well, it is assumed that the data indicating the trajectories E1 ′, E2 ′, E1 ″, and E2 ″ obtained as a result of the interpolation includes the difference values from the original template data values E1 and E2.

Rule 3: Interpolation processing for smoothing over the entire section of the element is performed. This example is shown in FIG.
Shown in Without changing the template (envelope waveform) AE1 of the first element and the template (envelope waveform) AE3 of the third element, the entire data of the template (envelope waveform) AE2-b of the second element in the middle is obtained. Interpolation is performed such that the leading end matches the end of the template (envelope waveform) AE1 of the first element and the end matches the beginning of the template (envelope waveform) AE3 of the third element. In this case as well, the data E2 'obtained as a result of the interpolation is assumed to be composed of a difference value (with a plus or minus sign) from the original template value (envelope value) E2. A specific example of the interpolation process of Rule 3 is as follows.
As shown in FIGS. 20A, 20B and 20C, there are a plurality of variations. FIG. 20A shows an example in which interpolation is performed using only the intermediate element AEn. E1 indicates the original trajectory of the template data value of the element AEn. The locus E1 of the template data value of the element AEn is shifted according to the difference between the template data value EP0 of the end point of the preceding element AEn-1 and the template data value SP of the original start point of the intermediate element AEn. , The template data including the trajectory Ea is created for all the sections of the element AEn. The locus E1 of the template data value of the element AEn is shifted according to the difference between the template data value EP of the original end point of the intermediate element AEn and the template data value SP1 of the start point of the succeeding element AEn + 1. Then, template data including the trajectory Eb is created corresponding to all sections of the element AEn. Next, the template data of the trajectory Ea and the template data of the trajectory Eb are cross-fade-interpolated so as to smoothly change from Ea to Eb. Get it.

FIG. 20B shows an intermediate element AEn.
In this example, the data is changed in all the sections and the interpolation is performed in the predetermined area RCFT at the end of the intermediate element AEn and the predetermined area FCFT at the end of the succeeding element AEn + 1. First, similarly to the above, the template data value E at the end point of the preceding element AEn-1
According to the difference between P0 and the template data value SP of the original starting point of the intermediate element AEn, the element AEn
Is shifted, and template data composed of the locus Ea is created corresponding to all sections of the element AEn.

Next, the intermediate level MPa between the template data value EPa at the end point of the trajectory Ea and the template data value SP at the start point of the succeeding element AEn + 1 is set as a target value, and the predetermined level at the end of the preceding element AEn is determined. In the area RCFT, the trajectory E of the preceding element AEn
Interpolation is performed so that the template data value of a approaches the target value MPa. As a result, the locus Ea of the template data of the preceding element AEn is shifted from the original locus by Ea ′.
Changes as shown. Further, in a predetermined area FCFT at the leading end of the succeeding element AEn + 1, the template data value of the succeeding element AEn + 1 is changed to the intermediate value MPa.
, And interpolation is performed so as to approach the original locus of the template data value indicated by the line E2. as a result,
The locus of the template data value of the succeeding element AEn + 1 in the interpolation area FCFT gradually approaches the original locus E2 as shown by the line E2 '.

FIG. 20C shows an intermediate element AEn.
Data is changed in all sections of
Interpolation is performed between a predetermined area RCFT at the end of En-1 and a predetermined area FCFT at the front end of the intermediate element AEn, and a predetermined area RCFT at the end of the intermediate element AEn and the front end of the subsequent element AEn + 1 are interpolated. An example is shown in which interpolation is performed with a predetermined area FCFT.
First, the original trajectory E1 of the template data value of the intermediate element AEn is shifted by an appropriate offset amount OFST, and template data including the trajectory Ec is created corresponding to all the sections of the element AEn.

Next, in the predetermined area RCFT at the end of the preceding element AEn-1 and the predetermined area FCFT at the tip of the intermediate element AEn, the interpolation processing is performed so that the trajectories E0 and Ec of both template data are smoothly connected. To obtain trajectories E0 'and Ec' as interpolation results in the interpolation area. Also, the intermediate element AEn
A predetermined area RCFT at the end portion of
Interpolation processing is performed so that the trajectories Ec and E2 of the two template data are smoothly connected to each other in the predetermined area FCFT at the leading end portion of FIG.
And E2 ″ in the interpolation area. Also in FIG. 20, each of the trajectories E1 ', Ea, Ea', E
Data indicating 2 ′, Ec, Ec ′, Ec ″, E0 ′
It is assumed that it consists of a difference value from the original template data values E1, E2, E0.

[Conceptual explanation of tone synthesis processing including connection processing] FIG. 21 performs the above-described connection processing for each template data corresponding to each tone element, and performs tone synthesis processing based on the connected template data. FIG. 3 is a block diagram conceptually illustrating the configuration of the musical sound synthesizer as described above.
Template data supply blocks TB1, TB2, TB
3 and TB4, waveform template data Ti relating to the preceding articulation element, respectively.
mb-Tn, amplitude template data Amp-Tn, pitch template data Pit-Tn, time template data TSC-Tn, and waveform template data Tim for the following articulation element
b-Tn + 1, amplitude template data Amp-Tn + 1,
The pitch template data Pit-Tn + 1 and the time template data TSC-Tn + 1 are supplied.

Rule data processing blocks RB1, R
In B2, RB3, and RB4, the connection rule TimR for each musical sound element related to the articulation element is used.
ULE, AmpRULE, PitRULE, and TSCRULE are decoded, and the connection processing as described with reference to FIGS. 17 to 20 is executed according to the decoded connection rules. For example, the rule decoding process block RB1 for the waveform template performs a process for executing the connection process (direct connection or cross-fade interpolation) described with reference to FIG.

The rule template decoding block RB2 for the amplitude template performs a process for executing the connection process (direct connection or interpolation) described with reference to FIGS. In this case, the interpolation result is given as a difference value (with a plus or minus sign) as described above. Therefore, the interpolation data composed of the difference value output from the block RB2 is originally supplied from the template data supply block TB2 in the adder AD2. Are added to the template data values of the above. For the same reason, each output of the other rule data processing blocks RB3 and RB4 and each template data supply block TB3 and TB4
Adders AD3 and AD4 for respectively adding the original template data values supplied from.

Thus, each of the adders AD2, AD3, AD
4, template data Amp, Pitch, and T, which are obtained by performing necessary connection processing between adjacent elements.
SC are output respectively. Pitch control block CB3
Controls the waveform reading speed according to the pitch template data Pitch. Since the waveform template itself contains the original pitch information, the original pitch information (original pitch envelope) is received from the database via the line L1, and the waveform is calculated based on the deviation between the original pitch envelope and the pitch template data Pitch. Control the reading speed. For example, when the original pitch envelope and the pitch template data Pitch are the same, reading may be performed at a constant waveform reading speed, and the original pitch envelope and the pitch template data P may be read.
If it is different from "itch", the waveform reading speed may be variably controlled by the deviation. Further, the pitch control block CB3 receives note instruction data, and controls the waveform reading speed based on the note instruction data. For example, assuming that the original pitch of the waveform template data is based on the pitch of the note C4 and that the sound of the note D4 is also generated using the waveform template data having the original pitch of the note C4, The waveform reading speed is controlled according to the deviation between the note D4 and the note C4 having the original pitch. The details of such pitch control will not be described in detail because a known technique can be applied.

Basically, the waveform access control block CB1 sequentially reads out each sample of the waveform template data in accordance with the waveform readout speed control information output from the pitch control block CB3. At this time, the waveform reading mode is controlled according to the TSC control information provided as the time template data, and the pitch of the generated sound is determined according to the waveform reading speed control information provided from the pitch control block CB3, while the total waveform reading time is TS
Variable control is performed according to the C control information. For example, if the sounding time length is to be longer than the time length of the original waveform data, the waveform reading speed is left as it is,
If some waveforms are read out in duplicate,
The sounding time length can be extended while maintaining a desired pitch. Also, when compressing the sounding time length rather than the time length of the original waveform data, the desired pitch can be maintained if the waveform reading speed is kept as it is and some of the waveform portions are skipped and read. The length of the sounding time can be compressed while performing. Waveform access control block CB
1 and the crossfade control block CB2 perform processing for executing the connection processing (direct connection or crossfade interpolation) described with reference to FIG. 17 according to the output of the waveform template rule decoding block RB1. Do. The crossfade control block CB2 is also used when performing crossfade processing while loop-reading a partial waveform template according to the partial vector PVQ. It is also used for smoothing the waveform connection during the TSC control.

The amplitude control block CB4 gives an amplitude envelope corresponding to the amplitude template Amp to the generated waveform data. Also in this case, since the waveform template itself contains the original amplitude envelope information, the original amplitude envelope information is received from the database via the line L2, and the original amplitude envelope and the amplitude template data Am are received.
The amplitude of the waveform data is controlled by the deviation from p. For example, when the original amplitude envelope and the amplitude template data Amp are the same, the amplitude control block CB4 only needs to pass the waveform data without performing substantial amplitude control. If the original amplitude envelope is different from the amplitude template data Amp, the amplitude level may be variably controlled by the deviation.

[Specific Example of Tone Synthesizing Apparatus] FIG. 22 is a block diagram showing an example of a hardware configuration of a tone synthesizing apparatus according to an embodiment of the present invention. This musical sound synthesizer may be an electronic musical instrument, a karaoke device, an electronic game device, another multimedia device, a personal computer, or the like.
It may take any product application form. According to the configuration shown in FIG. 22, the tone synthesis processing according to the embodiment of the present invention is executed using the software sound source. A software system is constructed so as to realize the tone data creation and tone synthesis processing according to the present invention, and a required database DB is constructed in an attached memory device, or communicates with a database DB constructed outside (host). An embodiment of accessing via a line is taken.

In the musical tone synthesizing apparatus shown in FIG. 22, a CPU (central processing unit) 10 is used as a main control unit, and under the control of the CPU 10, the generation of musical tone data and the musical tone synthesizing process according to the present invention are realized. The software program is executed, and the software sound source program is executed. Of course, the CPU 10 can also execute other appropriate programs in parallel. CP
U10 has a ROM (Read Only Memory) 11, R
AM (random access memory) 12, hard disk device 13, first removable disk device (for example, C
D-ROM drive or MO drive) 14, second
(Eg, a floppy disk drive) 15, a display 16, an input operation device 17, such as a keyboard and a mouse, a waveform interface 18, a timer 19, a network interface 20, a MIDI interface 21, and the like via a data and address bus 22. Connected.

FIG. 23 shows a detailed example of the waveform interface 18 and a configuration example of the waveform buffer in the RAM 12. The waveform interface 18 controls both acquisition (sampling) and output of waveform data.
An analog / digital converter (ADC) 23 for sampling waveform data inputted from outside by a microphone or the like and performing analog / digital conversion; a first DMAC (direct memory access controller) 24 for sampling; And a second DMAC for controlling the output of waveform data.
(Direct memory access controller) 26 and a digital / analog converter (DAC) 27 for digital / analog conversion of output waveform data. Note that the second DMAC 26 has a sampling clock Fs
, And provides the absolute time information to the bus 22 of the CPU.

The RAM 12 has a plurality of waveform buffers W-BUF. One waveform buffer W-BUF
Has a storage capacity (number of addresses) for storing waveform sample data for one frame. For example, when the reproduction sampling frequency based on the sampling clock Fs is 48 kHz
z, assuming that the time of one frame section is 10 milliseconds, one waveform buffer W-BUF has a capacity to store 480 samples of waveform sample data. When at least two waveform buffers W-BUF (A, B) are used and one waveform buffer W-BUF is in read mode and accessed by DMAC 26 of waveform interface 18, the other waveform buffer W-BUF is written. Mode and writes the generated waveform sample data. In the tone synthesis program according to this embodiment, waveform sample data composed of a plurality of samples for one frame is collectively generated for each tone synthesis channel, and one waveform buffer W in the write mode is generated.
-The waveform sample data of each channel is added (accumulated) to each sample position (address position) of BUF. For example, if one frame is composed of 480 samples, waveform sample data of 480 samples for the first tone synthesis channel is collectively calculated,
This is stored at each sample position (address position) of the waveform buffer W-BUF. Next, 480 samples of waveform sample data for the second tone synthesis channel are collectively operated, and this is used as the same waveform buffer W-.
Each sample position (address position) of the BUF is added (accumulated). Hereinafter, the same applies. Therefore, when the generation calculation of the waveform sample data for one frame for all the channels is completed, the waveform samples of all the channels are stored in each sample position (address position) of one waveform buffer W-BUF in the write mode. Total waveform sample data obtained by accumulating data for each sample is stored. For example, first, the total waveform sample data for one frame is written into the waveform buffer W-BUF of A, and then the total waveform sample data of one frame is written into the waveform buffer W-BUF of B. A-waveform buffer W-BUF
As soon as writing is completed, the mode shifts to the reading mode from the beginning of the next frame section, and is read out regularly during the frame section at a predetermined reproduction sampling cycle based on the sampling clock Fs. Therefore, basically, the read and write modes of the two waveform buffers W-BUF (A, B) may be alternately switched and used. However, if there is a margin for writing ahead several frames ahead, Three or more waveform buffers W-BUF (A, B,
C, ...) may be used.

Under the control of the CPU 10, a software program for creating tone data and tone synthesis processing according to the present invention is stored in the ROM 11, the RAM 12, the hard disk device 13, or the removable disk devices 14, 15. You may do so. Further, it is connected to a communication network via the network interface 20 and, from an external server computer (not shown), the above-mentioned "program for realizing tone data creation and tone synthesis processing according to the present invention", data in a database DB, etc. To the internal RAM 12
Alternatively, it may be stored in the hard disk 13 or the removable disk devices 14, 15, or the like. CPU10
Executes, for example, a “program for realizing tone data creation and tone synthesis processing according to the present invention” stored in the RAM 12 to synthesize a tone according to a rendition style sequence, and converts the synthesized tone waveform data into a waveform in the RAM 12. The data is temporarily stored in the buffer W-BUF. Under the control of the DMAC 26, waveform data is read out from the waveform buffer W-BUF in the RAM 12 and read from the digital / analog converter (D
AC) 27 for D / A conversion. The D / A converted musical sound waveform data is applied to a sound system (not shown) and spatially generated.

As shown in FIG. 8A, the following description is based on the assumption that the data of the rendition style sequence (articulation element sequence AESEQ) according to the present invention is incorporated in the automatic performance sequence data composed of MIDI data. Do. Although not described in detail in FIG. 8A, the data of the rendition style sequence (articulation element sequence AESEQ) can be incorporated in the form of MIDI format, for example, as exclusive data of MIDI.

FIG. 24 is a time chart showing an outline of a musical sound generation process executed by a software tone generator based on performance data in the MIDI format. (A)
The “performance timing” shown in FIG. 8 indicates a MIDI note-on event, a note-off event or another event (EVENT (MIDI) in FIG. 8A), and an articulation element sequence event (EVENT (MIDI) in FIG. 8A). AESEQ)) and other events # 1 to # 4. (B) illustrates the relationship between the timing at which waveform sample data is generated and calculated (“waveform generation”) and its reproduction timing (“waveform reproduction”). The “Waveform Generation” column at the top shows each sample of one waveform buffer W-BUF that is in a write mode by collectively generating waveform sample data composed of a plurality of samples for one frame for each tone synthesis channel. The timing at which the process of adding (accumulating) the waveform sample data of each channel to the position (address position) is illustrated. The lower column of “waveform reproduction” indicates the timing at which processing for regularly reading waveform sample data from the waveform buffer W-BUF at a predetermined reproduction sampling cycle based on the sampling clock Fs during one frame period. The indications of A and B attached to the waveform buffers W to be written or read out respectively.
-A symbol that distinguishes what BUF is. FR
1, FR2, FR3,... Are provisionally assigned frame numbers. For example, the waveform sample data for a certain frame subjected to the waveform generation operation in the frame FR1 is written into the waveform buffer W-BUF of A, and this is written into the waveform buffer W-BU of A in the next frame FR2.
Read from F. The waveform sample data for the next one frame is generated and calculated in frame FR2.
Is written to the waveform buffer W-BUF. The waveform sample data for one frame stored in the waveform buffer W-BUF for B is further stored in the next frame FR3.
From the waveform buffer W-BUF. Events # 1, # 2, and # 3 shown in (a) occur within the time of one frame, and the generation calculation of the waveform sample data corresponding to these events # 1, # 2, and # 3 is (b) ) In frame FR3. Therefore, the rise (start of sound generation) of the musical tone corresponding to these events # 1, # 2, and # 3 is started in the next frame FR4. Δt indicates the difference between the timing of occurrence of the events # 1, # 2, and # 3 given as MIDI performance data and the timing of starting the generation of the corresponding musical tone. Since this time shift Δt is only for one to several frames, there is no problem in audibility. Note that the waveform sample data at the start of sound generation is not written from the beginning of the waveform buffer W-BUF, but the waveform buffer W corresponding to the start time is written.
-BUF is written from a predetermined halfway position.

[0116] The method of generating and calculating the waveform sample data in the "waveform generation" includes an automatic performance sound based on a normal MIDI note-on event (this is called a "normal performance" sound) and an articulation sound. With the performance sound based on the ON event of the element sequence AESEQ (this is referred to as a “performance style performance” sound),
Is different. “Normal performance” processing based on a normal MIDI note-on event and “performance technique” processing based on an articulation element sequence AESEQ on event are performed as shown in FIGS. 29 and 30.
Each is executed by a separate processing routine. For example, if the accompaniment part is performed by "normal performance" based on a normal MIDI note-on event, and a specific solo performance part is performed by "performance technique" based on the articulation element sequence AESEQ,
It is effective.

FIG. 25 shows a rendition style sequence (articulation element sequence AES) according to the present invention.
6 is a time chart showing an outline of a “playing style performance” process (musical sound synthesis process of an articulation element) based on EQ) data. The “phrase preparation command” and the “phrase start command” are included in the MIDI performance data as “articulation element sequence event EVENT (AESEQ)” as shown in FIG. 8A. That is, 1
Articulation element sequence AE
The event data of SEQ (referred to as “phrase” in FIG. 25) includes a “phrase preparation command” and a “phrase start command”. The preceding event data, the “phrase preparation command”, includes an articulation element sequence AES to be reproduced.
It designates an EQ (that is, a phrase) and indicates that preparation for playback is to be performed, and is given a predetermined time earlier than the sounding start time of the articulation element sequence AESEQ. Block 3
In the “preparation processing” process indicated by “0”, all data necessary for reproducing the specified articulation element sequence AESEQ is retrieved from the database DB according to the “phrase preparation command”, Downloaded to the buffer area and the articulation element sequence A
Necessary preparations are made so that the ESEQ is developed and the articulation element sequence can be reproduced immediately. Also, in this "preparation process" process,
It also performs processing for interpreting the specified articulation element sequence AESEQ, setting or determining a rule or the like for connecting successive articulation elements, and forming necessary connection control data and the like.
For example, assuming that the designated articulation element sequence AESEQ is composed of five articulation elements AE # 1 to AE # 5 as shown in the figure, the connection points (points indicated as connection 1 to connection 4) at each connection point The connection rule is determined, and connection control data for the connection rule is formed. Further, data indicating the start time of each of the articulation elements AE # 1 to AE # 5 is prepared in a relative time expression from the start of the phrase. A “phrase start command”, which is event data subsequent to the “phrase preparation command”, instructs the start of sounding of the articulation element sequence AESEQ. In response to the "phrase start command", each of the articulation elements AE # 1 to AE # prepared in the "preparation process"
5 are reproduced sequentially. That is, when the start time of each of the articulation elements AE # 1 to AE # 5 arrives, the corresponding articulation element AE #
The playback of the first articulation element AE # 1 is started smoothly at each connection point (connection 1 to connection 4) according to the prepared connection control data. A predetermined connection process is performed so as to be connected.

FIG. 26 is a flowchart showing a main routine of a tone synthesis process executed by the CPU 10 of FIG. By the "automatic performance process" of the main routine, a process based on the event of the automatic performance sequence data is performed. First, in step S50, various necessary initialization processes such as securing various buffer areas on the RAM 12 are performed. Next, in step S51, it is checked whether or not each of the following activation factors has occurred. Activation factor: MI via interface 20, 21
DI performance data or other communication input data is input. Activation factor: Automatic performance processing timing has arrived.
In order to check the time of occurrence of the next event in the automatic performance, the automatic performance processing timing occurs regularly. Activation factor: Waveform generation timing of one frame has arrived. In order to generate the waveform sample data collectively in units of one frame, the waveform generation timing is generated in one frame cycle (for example, at the end of a frame section). Activation factor: A switch operation of a keyboard or a mouse (except for a main routine end instruction operation) performed by the input operation device 17. Activation factor: Disk drives 13 to 15 and display 16
Interrupt request from Activation factor: The end instruction operation of the main routine is performed by the input operation device 17.

In step S52, it is determined whether or not any of the activation factors has occurred. If NO, steps S51 and S52 are repeated, and if YES, it is determined in step S53 which activation factor has occurred. When the activation factor has occurred, a predetermined “communication input process” is performed in step S54. If an activation factor has occurred, a predetermined "automatic performance process" (an example of which is shown in FIG. 27) is performed in step S55. If an activation factor has occurred, a predetermined "sound source process" is executed in step S56 (an example of which is shown in FIG. 2).
8). If an activation factor has occurred, a predetermined "SW process" (process corresponding to the operated switch) is performed in step S57. If an activation factor has occurred, predetermined "other processing" (processing corresponding to the interrupt request) is performed in step S58. If an activation factor has occurred, a predetermined "end processing" (processing for ending this main routine) is performed in step S59.

In step S53, if it is determined that two or more activation factors among the activation factors are simultaneously occurring, a predetermined priority order (eg, activation factors,...,. In that order). In that case, there may be equal priority processing. Steps S51 to S53 virtually show task management in the pseudo multitasking process. Actually, while the process is being executed based on the occurrence of one of the activation factors, the process starts. Also execute another process by an interrupt due to the occurrence of an activation factor with a higher priority (for example, if an activation factor occurs during execution of "sound source processing" based on the occurrence of an activation factor) To execute “automatic performance processing” by interruption).

A specific example of the "automatic performance process" (step S55) will be described with reference to FIG. First, step S
At 60, a process of comparing the absolute time information given from the DMAC 26 (FIG. 23) with the next event timing of the music data is performed. As shown in FIG. 8, in the music data, that is, in the automatic performance data, the event data EVENT
Precedes the duration data DUR. For example, when the duration data DUR is read out, the absolute time information at that time is added to the duration data DUR to create and store absolute time information indicating the arrival time of the next event. Then, the absolute time information indicating the arrival time of the next event is compared with the absolute time information at the present time in step S60 in FIG.

In step S61, it is determined whether or not the current absolute time coincides with or has passed the next event arrival time. If the next event has not yet arrived, Figure 2
The process of 7 is immediately terminated. When the next event arrives, the process goes to step S62, and the type of the event is a normal performance event (ie, a normal MIDI event) or a performance style event (ie, an articulation element sequence event). Find out if there is. If it is a normal performance, go to step S63,
Normal MIDI event processing corresponding to the event is performed to generate sound source control data. In the next step S64, a tone synthesis channel (abbreviated as "sound source channel" in the figure) relating to the event is detected, and the number of the channel is registered in the channel number register i. For example, in the case of a note-on event, a channel to which the occurrence of the note is assigned is determined, and the channel is registered in the register i. In the case of a note-off event, the channel to which the occurrence of the note is assigned is detected,
The channel is registered in the register i. Next step S
In step 65, the tone buffer TBUF (i) of the channel number designated by the register i is stored in step S63.
The sound source control data and the control timing data generated in step (1) are stored. Note that the control timing is a timing at which control relating to the event is performed, such as a sound generation start timing for a note-on event and a release start timing for a note-off event. In this embodiment, since the tone waveform is generated by software processing, the timing of the MIDI data event occurrence and the timing of the actual processing corresponding thereto are slightly shifted. Thus, the actual control timing is instructed again.

If it is determined in step S62 that the event is a performance technique event, the flow advances to step S66 to check whether the event is a "phrase preparation command" or a "phrase start command" (see FIG. 25). If it is a "phrase preparation command", steps S67 to S7
1 is executed. Steps S67 to S71
This routine corresponds to the “preparation process” indicated by the block 30 in FIG. First, in step S67, a tone synthesis channel (abbreviated as "sound source channel" in the figure) for reproducing the phrase (that is, the articulation element sequence AESEQ) is determined, and the channel number is registered in the register i. In the next step S68, a performance style sequence (abbreviated as “performance style SEQ” in the figure) of the phrase (that is, articulation element sequence AESEQ) is developed. That is, the articulation element sequence AESEQ is decomposed to the level of vector data capable of indicating an individual template, analyzed, and connected to each articulation element (AE # 1 to AE # 5 in FIG. 25). The connection rules for 1 to 4) are determined, and connection control data therefor is formed. In step S69, it is checked whether or not there is a subsequence (abbreviated as "subSEQ" in the figure). If there is, the process returns to step S68 to further decompose the subsequence to a level of vector data capable of indicating an individual template.

FIG. 32 shows an example in which the articulation element sequence AESEQ includes a subsequence. As shown in FIG. 32, the articulation element sequence AESEQ may have a hierarchical structure. That is, in the figure, “performance style SEQ # 2” is M
If it is specified by the data of the articulation element sequence AESEQ incorporated in the IDI performance information, the specified sequence “performance style SEQ # 2” is composed of “performance style SEQ # 6” and “element vector E-VEC # 5 ". This “performance style SEQ # 6” corresponds to a subsequence. By analyzing this sub-sequence, “performance style SEQ # 6” is specified by element vectors E-VEC # 2 and E-VEC # 3. Thus, MI
“Reproduction style SEQ # 2” specified by the data of the articulation element sequence AESEQ incorporated in the DI performance information is developed, and this is expressed by element vectors E-VEC # 2, E-VEC # 3, E-VEC # 3.
It is analyzed that it is specified by VEC # 5. As described above, at this time, connection control data for connecting each articulation element is also formed as needed. The element vector E-VEC is data that specifically specifies an individual articulation element. Of course,
Not only in the case of having such a layered structure, but also from the beginning, each element vector E-E is generated by the “performance style SEQ # 2” specified by the data of the articulation element sequence AESEQ incorporated in the MIDI performance information. VEC # 2, E-VEC # 3, E-VE
In some cases, C # 5 is specified.

In step S70, the data of each expanded element vector (abbreviated as "E-VEC" in the figure) together with the data indicating the control timing of the element vector by the relative time, together with the tone buffer TBUF of the channel number designated by the register i. (I). In this case, the control timing is the start timing of each articulation element as shown in FIG. In the next step S71, the tone buffer TBUF
Referring to (i), necessary template data is loaded from the database DB to the RAM 12. If the current event is a “phrase start command” (see FIG. 25), the routine of steps S72 to S74 is executed. In this step S72, the channel to which the phrase performance is reproduced is detected, and the channel number is registered in the register i. In the next step S73, all control timing data stored in the tone buffer TBUF (i) of the channel number designated by the register i is converted into data in absolute time expression. That is, the absolute time information given from the DMAC 26 when the "phrase start command" is generated is used as an initial value, and the initial value is added to the relative time of each control timing data, so that each control timing data is converted to the absolute time. Can be converted to expression data. In the next step S74, the tone buffer TBUF
The content of (i) is rewritten according to the absolute time of each converted control timing. That is, the start time and end time of each element vector E-VEC constituting the rendition style sequence, connection control data between the element vectors, and the like are written in the tone buffer TBUF (i).

Next, referring to FIG. 28, "sound source processing" (FIG.
A specific example of step S56) will be described. As described above, the “sound source processing” is activated for each frame.
First, in step S75, a predetermined waveform generation preparation process is performed. For example, the contents of the waveform buffer W-BUF that has been reproduced and read in the previous frame section are cleared, and data can be written to the waveform buffer W-BUF in the current frame section. In the next step S76, it is checked whether or not there is a channel for which sound generation processing is to be performed. If not, the process does not need to be continued, and the process jumps to step S83. if there is,
In step S77, one of the channels to be subjected to sound generation processing is specified, and preparations are made for performing waveform sample data generation processing for the channel. In the next step S78, it is checked whether the type of musical tone assigned to the prepared channel is a "normal performance" sound or a "performance style" sound. If the sound is a "normal performance" sound, the process proceeds to step S79, and the waveform sample data for one frame for the channel is converted to the "normal performance" sound.
A process for generating a sound is performed. If the sound is a “performance style” sound, the process proceeds to step S80, and a process of generating one frame of waveform sample data for the channel as a “performance style performance” sound is performed. Next, in step S81, it is checked whether or not there is a remaining (unprocessed) channel among the channels to be subjected to the sound generation processing. If there is, the process goes to step S82 to specify the next channel to be processed from the remaining (unprocessed) channels, and prepares to perform the waveform sample data generation process for the channel. Then, the process returns to the step S78, and the same processing of the steps S78 to S80 as described above is executed for a new channel. When the processes of steps S78 to S80 are completed for all the channels for which the sound generation process is to be performed, there is no remaining (unprocessed) channel, so that step S81 is NO and the process proceeds to step S83. In this state, the generation of one frame of waveform sample data for all the channels to be sounded is completed, and they are added (accumulated) for each sample and stored in the waveform buffer W-BUF. . In step S83, the data in the waveform buffer W-BUF is delivered under the control of the waveform input / output (I / O) driver. Thus, in the next one frame period, the waveform buffer W-BUF
Is in the read mode, and is accessed by the DMAC 26, and the waveform sample data is reproduced and read at a regular sampling cycle according to a predetermined sampling clock Fs.

FIG. 29 shows a detailed example of the process in step S79 in FIG. FIG. 29 is a flow chart showing an example of "1 frame of waveform data generation process" for "normal performance", in which a normal tone synthesis process based on MIDI performance data is performed. In this process, each time the loop of steps S90 to S98 is performed once, one sample of waveform data is generated. Therefore, address pointer management indicating the number of a sample being processed in one frame is performed, but this is not described in detail. First, in step S90, it is checked whether the control timing has arrived. This control timing is the timing specified again in step S65 in FIG. 27, and is, for example, a sound generation start timing or a release start timing (sound generation end timing). If there is any control timing for the frame currently being processed, step S9 corresponds to the address pointer value corresponding to the time of the control timing.
If the answer is 0, the process goes to step S91 to perform a necessary waveform generation start process based on the sound source control data. If the current address pointer value does not correspond to the control timing, step S91 is jumped to step S92. In step S92, a process for forming a low-frequency signal (LFO) necessary for vibrato or the like is performed. Next step S
At 93, a process for forming an envelope signal (EG) for pitch control is performed.

In the next step S94, based on the sound source control data, waveform sample data of a predetermined timbre is stored in a waveform memory (not shown) for a "normal performance" sound at a rate corresponding to the designated tone pitch. And performs a process of interpolating the value of the read waveform sample data between samples. Here, a commonly known waveform memory reading technique and inter-sample interpolation technique may be appropriately used. The tone pitch specified here is the regular pitch of the note (pitch) related to the note-on event, and
This is variably controlled by the vibrato signal, pitch control envelope value, and the like formed at 93. In the next step S95, a process for forming an amplitude envelope (EG) is performed. In the next step S96, step S94
The volume level of the waveform data of one sample generated in step S95 is variably controlled by the amplitude envelope value formed in step S95, and this is already stored at the address of the waveform buffer W-BUF indicated by the current address pointer. Add to the waveform sample data. That is, the addition and accumulation are performed on the waveform sample data of another channel for the same sample point. Next, in step S97, it is determined whether or not processing for one frame has been completed. If not completed, the process goes to step S98 to prepare the next sample (the address pointer is advanced to the next).

With the above configuration, when sound generation is started in the middle of a frame, waveform sample data is stored from an intermediate address of the waveform buffer W-BUF corresponding to the sound generation start position. Of course, when sound generation is continued throughout one frame period, the waveform sample data is stored in all addresses of the waveform buffer W-BUF. Note that the envelope forming processing in steps S93 and S95 may be performed by reading an envelope waveform memory, or may be performed by calculating a predetermined envelope function.
As the envelope function, a well-known method of calculating a relatively simple linear function of the first order may be used. Note that, unlike “performance technique performance” to be described later, in “normal performance”, complicated processing such as replacement of a sounding waveform, replacement of an envelope, and control of time axis expansion / contraction of a waveform need not be performed.

FIG. 30 shows a detailed example of the processing in step S80 in FIG. FIG. 30 is a flowchart showing an example of "processing for generating waveform data for one frame" for "performance technique", in which an articulation (performance technique) is performed.
A tone synthesis process based on the sequence data is performed here. In the process of FIG. 30, the tone waveform processing of the articulation element based on each template data, the connection processing between element waveforms, and the like are executed in the manner already described. Similarly to FIG. 29, in the process of FIG. 30, each time the loop of steps S100 to S108 is performed once, waveform data of one sample is generated. Therefore, address pointer management indicating the number of a sample being processed in one frame is performed, but this is not described in detail. In the processing of FIG. 30, in order to smoothly connect the articulation elements that are adjacent to each other, two-series various template data (including the waveform template) are cross-fade synthesized, and the time series expansion / contraction control is performed. Cross-fade synthesis of two series of waveform sample data is performed.
Therefore, various data processing for two series for crossfade synthesis is performed for one sample point.

First, in step S100, it is checked whether the control timing has arrived. This control timing is the timing written in step S74 of FIG. 27, and is, for example, the start timing of each of the articulation elements AE # 1 to AE # 5, the start timing of the connection processing, and the like. If there is any control timing with respect to the frame currently being processed, step S100 is YES in response to the address pointer value corresponding to the time of the control timing, and the process goes to step S101 to execute the element corresponding to the control timing. Necessary control is performed based on the vector E-VEC, connection control data, and the like. If the current address pointer value does not correspond to the control timing, step S101 is jumped to step S102.

In step S102, a time template for a specific element specified by the element vector E-VEC (the template is set to TM
P) (abbreviated as P). The time template is the time template (TSC template) shown in FIG. In this embodiment, the time template (TSC template) is provided as envelope-like data that changes over time, like the amplitude template and the pitch template. Therefore, in this step S102, a process of forming an envelope of the time template is performed. Step S103
Then, a process of generating a pitch template for a specific element specified by the element vector E-VEC is performed. The pitch template is also given as time-varying envelope-shaped data as illustrated in FIG. In step S105, a process of generating an amplitude (Amp) template for a specific element specified by the element vector E-VEC is performed. The amplitude template is also given as envelope-like data that changes with time as illustrated in FIG.

Steps S102, S103, S105
In the same manner as described above, the envelope forming method may be performed by reading an envelope waveform memory, or may be performed by calculating a predetermined envelope function. A relatively simple method of calculating a first-order linear function may be used. Also, as described with reference to FIGS. 18 to 20, templates are formed in two series (templates of preceding elements and templates of succeeding elements) corresponding to predetermined element connection locations, and both are connected and controlled. In steps S102, S103, and S105, a process of connecting by cross-fade synthesis according to the data and an offset process are also performed. Which connection rule is used to perform the connection process differs depending on the corresponding connection control data.

In step S104, basically, a waveform (Timbre) template for a specific element specified by the element vector E-VEC is read out at a rate corresponding to the specified tone pitch. The tone pitch specified here is the same as the previous step S1
It is variably controlled by the pitch template (pitch control envelope value) formed in step 03.
In step S104, control for expanding or compressing the existence time of the waveform sample data along the time axis, that is, TSC control, is performed in accordance with the time template (TSC template) independently of the musical tone pitch. In addition, the waveform sample data (two waveform sample data corresponding to different time points in the same waveform template) are read out in two series so that the continuity of the waveform is not impaired by the time axis expansion / contraction control. Is also performed in step S104. In addition, as in the case of “normal performance”, the interpolation calculation process between waveform samples is also performed in step S104. Further, as described with reference to FIG. 17, the waveform templates are read in two series (corresponding to the predetermined element connection locations) (the waveform template of the preceding element and the waveform template of the following element), and the two are cross-fade combined. The connection process is also performed in step S104. Further, FIG.
As described with reference to FIGS. 3 to 16, the process of loop-reading (repeatingly reading) the waveform template and the process of cross-fading the two series of loop-read waveforms at this time are also performed in step S104. In addition,
When a waveform (Timbre) template to be used keeps the temporal pitch fluctuation component in the original waveform as it is, the value of the pitch template is given by the amount of change (difference value or ratio) with respect to the original pitch fluctuation. Good to do. In other words, when the original temporal pitch fluctuation is to be kept as it is, the value of the pitch template is maintained at a constant value (for example, “1”).

In the next step S105, processing for forming an amplitude template is performed. In the next step S106, the volume level of the waveform data of one sample generated in step S104 is variably controlled by the amplitude envelope value formed in step S105, and this is controlled by the waveform buffer W-BUF indicated by the current address pointer. Add to the waveform sample data already stored at the address location. That is, the addition and accumulation are performed on the waveform sample data of another channel for the same sample point.
Next, in step S107, it is determined whether or not processing for one frame has been completed. If not completed, the process goes to step S108 to prepare the next sample (the address pointer is advanced to the next). As described above, when the waveform (Timbre) template to be used retains the temporal amplitude fluctuation component in the original waveform as it is, the value of the amplitude (Amp) template is the amount of change with respect to the original amplitude fluctuation. (Difference value or ratio). That is, when the original temporal amplitude fluctuation is left as it is, the value of the amplitude template is maintained at a constant value (for example, “1”).

Next, an example of time axis expansion / contraction control (TSC control) will be described. High-quality multi-cycle waveform
In other words, with specific articulation characteristics,
Then, the waveform data consisting of a fixed amount of data (the number of samples or the number of addresses) is converted into the existence time length on the time axis independently of the tone reproduction pitch and without impairing the overall characteristics of the waveform. The arbitrarily variable control is performed by the present applicant in a separate application (for example, Japanese Patent Application No. 9-13039).
This can be realized by using the time axis expansion / contraction control (TSC control) proposed in No. 4). The point of the TSC control is as follows. A multi-period waveform composed of a constant waveform data amount is
In order to expand and contract the waveform data existence time length on the time axis while maintaining a constant reproduction sampling frequency and a predetermined reproduction pitch, in the case of compression, skip the appropriate part of the waveform data and read it out. In such a case, an appropriate portion of the waveform data is repeatedly read out,
Then, cross-fade synthesis is performed in order to remove discontinuity of waveform data due to jumping or partially repeated reading.

FIG. 31 is a diagram conceptually showing the outline of the time axis expansion / contraction processing (TSC control). (A) shows an example of a time template that changes with time. The time template is composed of data indicating a time axis expansion / contraction ratio (referred to as CRate). The vertical axis represents the data CRate, and the horizontal axis represents time t. The time axis expansion / contraction ratio data CRate indicates a ratio based on “1”,
A value of “1” indicates that the time axis is not expanded or contracted, a value larger than “1” indicates compression of the time axis, and a value smaller than “1” indicates expansion of the time axis. (B) to (d) of FIG.
Shows an example of performing time axis expansion / contraction control in accordance with the time axis expansion / contraction ratio data CRate using the virtual read address VAD and the real read address RAD. The solid line indicates the real read address RAD, and the broken line indicates the virtual read address VAD. (B)
Shows an example of time axis compression control according to the time axis expansion / contraction ratio data CRate (> 1) at point P1 in the time template of (a), and (c) shows the time axis compression Time axis expansion / contraction ratio data CRate
(D) shows an example in which the time axis does not expand or contract according to (= 1);
9A shows an example of time axis expansion control according to the time axis expansion / contraction ratio data CRate (<1) at point P3 in the time template of FIG. In (c), the solid line indicates the progress of the original waveform read address according to the pitch information, and the actual read address RAD and the virtual read address VAD
Matches.

The actual read address RAD is an address used for actually reading waveform sample data from the waveform template, and changes at a constant change rate according to desired pitch information. For example, by accumulating the frequency numbers corresponding to the desired pitch regularly, it is possible to obtain the actual read address RAD having a constant slope corresponding to the pitch. The virtual read address VAD assumes a state in which the desired length or the length of the waveform data on the time axis is controlled to be expanded or compressed, and in order to achieve the desired time axis expansion or compression, the waveform sample is determined from any address position at the present time. This address indicates whether data should be read. Therefore, desired pitch information and time axis expansion / contraction ratio data CRate
And the expansion ratio data C
Address data that changes at an inclination corrected by Rate is generated as a virtual read address VAD. Comparing the real read address RAD with the virtual read address VAD,
When the width of the real read address RAD separated from the virtual read address VAD exceeds a predetermined width, the real read address RAD
Is switched, and according to the switching instruction,
The numerical value of the real read address RAD is shift-controlled by an appropriate number of addresses so as to eliminate the deviation of the real read address RAD from the virtual read address VAD.

FIG. 33 is an enlarged view showing the same state as FIG. 31 (b). The dashed line exemplifies the original address progression according to the pitch information, and corresponds to the solid line in FIG. The thick broken line illustrates the address progress of the virtual read address VAD. Expansion ratio data CR
If ate is 1, the address advance of the virtual read address VAD matches the original address advance of the dashed line, and there is no change in the time axis. When compressing the time axis, the expansion / contraction ratio data CRate takes an appropriate value of 1 or more, and the inclination of the address progression of the virtual read address VAD becomes relatively large as shown in the figure. The bold solid line indicates the actual read address RAD
An example of the address progression will be described. This actual read address RAD
Of the address advance coincides with the original slope of the address advance according to the pitch information indicated by the dashed line. In this case, since the gradient of the address progress of the virtual read address VAD is relatively large, the address progress of the real read address RAD gradually changes with the elapse of time with the virtual read address VA.
It is later than the address advance of D. Then, when the width of the separation exceeds a predetermined value, a switching instruction (indicated by an arrow in the figure) is issued, and the actual read address RAD is shifted by an appropriate amount in the direction of eliminating the separation as shown in the figure. As a result, the address progression of the real read address RAD keeps the inclination according to the pitch information, while maintaining the inclination according to the pitch information.
The characteristic changes along the address progression of D and is compressed in the time axis direction. Therefore, such an actual read address R
By reading out the waveform sample data of the waveform template according to the AD, it is possible to obtain a waveform signal in which the waveform is compressed in the time axis direction without changing the pitch of the musical tone to be reproduced.

FIG. 34 is an enlarged view showing the same state as FIG. 31 (d). In this case, the expansion ratio data CRat
e is less than 1, and the inclination of the address progress of the virtual read address VAD indicated by the thick broken line is relatively small. Therefore, when the address progression of the real read address RAD gradually progresses with the lapse of time with respect to the address progression of the virtual read address VAD, and the width of the gap becomes a predetermined value or more,
A switching instruction (indicated by an arrow in the figure) is issued, and as shown in FIG.
Is shifted by an appropriate amount. As a result, the actual read address R
The address progression of the AD changes along the address progression of the virtual read address VAD while maintaining the inclination according to the pitch information, and exhibits a characteristic extended in the time axis direction. Therefore,
By reading the waveform sample data of the waveform template in accordance with such an actual read address RAD, it is possible to obtain a waveform signal whose waveform has been expanded in the time axis direction without changing the pitch of the musical tone to be reproduced.

The shift of the actual read address RAD in the direction to eliminate the separation should be such that the waveform data read immediately before the shift and the waveform data read immediately after the shift are smoothly connected by this shift. Is preferred. Further, as shown by a broken line in the drawing, it is preferable to perform crossfade synthesis in an appropriate period at the time of switching. The dashed line indicates the address progress of the cross-fade sub-system actual read address RAD2. As shown in the figure, when the above switching instruction is issued, the actual read address RAD2 for the cross-fade sub-system becomes the actual read address RA before shifting.
On the extension of the address progression of D, it is generated at the same rate (that is, gradient) as the actual read address RAD. During an appropriate cross-fade period, the sub-system actual read address RAD2
The cross-fade synthesis is performed so that the waveform smoothly transitions from the waveform read in response to the actual read address RAD for the main sequence to the waveform read in response to the actual read address RAD. In the case of this example, the sub-system actual read address RAD2 may be generated only at least during a required cross-fade period. It should be noted that the time axis expansion / contraction ratio CR is not limited to the TSC control example in which crossfade synthesis is partially performed as described above.
TSC control that always performs cross-fade synthesis processing in a mode corresponding to the value of ate may be employed.

In the case where waveform sample data is generated by repeatedly reading out the waveform template (that is, loop waveform) of the partial vector PVQ as shown in FIGS. 13 to 15, basically, the number of loops is varied. Thus, the time length of the entire loop readout waveform can be variably controlled relatively easily and independently of the tone reproduction pitch. That is, when a specific crossfade curve is specified by data specifying the crossfade section length, the crossfade section length (time length or number of loops) is determined accordingly. Here, the slope of the crossfade curve is variably controlled by the time axis expansion / contraction ratio indicated by the time template, so that the speed of the crossfade is variably controlled, and eventually the time length of the crossfade section is variably controlled. During that time, the tone reproduction pitch is not affected, so that the loop length is variably controlled, so that the time length of the cross-fade section is variably controlled.

When the existence time of the reproduced waveform data on the time axis is controlled by the time axis expansion / contraction control, it is desirable to control the time axis of the pitch template and the amplitude template in accordance with the expansion / contraction control. . Therefore, steps S103 and S105 in FIG.
In, according to the time template created in step S102, the time axis of the pitch template and the amplitude template created in this step is controlled to expand and contract.

It should be noted that all of the tone synthesis functions may not be constituted by software tone generators, but may be of a hybrid type of software tone generator and hardware tone generator. Also,
The tone synthesis processing according to the present invention may be performed only by the hardware tone generator. Alternatively, the tone synthesis processing according to the present invention may be performed using a DSP (Digital Signal Processor). When using a software tone generator, a hardware tone generator, or a hybrid tone generator, the waveform forming method is not limited to the simple PCM waveform memory reading method, but uses various data compression techniques as described above. An appropriate method can be used, such as a method based on the above, or a method based on parameter calculation according to various waveform synthesis algorithms.

[0145]

As described above, according to the present invention, a storage means for storing a plurality of waveform data dispersedly extracted from an original waveform to which vibrato is added, and storing one of the waveform data for a predetermined period of time. And a read unit that executes a predetermined read sequence by sequentially switching waveform data to be read, and realizes vibrato over a plurality of cycles by repeating this read sequence. Vibrato 1 of original waveform
By dispersively extracting and storing a plurality of waveform data from an appropriate range such as a cycle, a plurality of cycles, or less than one cycle, the storage capacity of the waveform data to be stored in the storage unit is reduced as compared with a case where the entire original waveform is stored. High quality vibrato waveform data that reflects vibrato characteristics in the original waveform as much as possible can be obtained by saving and by distributed extraction. As described above, since each waveform data extracted in a dispersed manner reflects the characteristic of vibrato in the original waveform, the musical tone waveform obtained by repeating and combining these waveform data is comparable to the vibrato in the original waveform. Since high-quality vibrato is included, and various ways of reading each waveform data from the storage means are variably controlled, various vibrato elements such as a vibrato cycle can be variously controlled. An excellent effect is obtained that the vibrato sound can be generated with rich controllability and the data storage capacity can be saved.

Further, by changing the reading speed of the waveform data over time or by changing the amplitude of the waveform data read out corresponding to the reading sequence over time,
It is possible to control the vibrato depth and the accompanying amplitude fluctuation. Further, the apparatus further comprises control data generating means for generating time control data for controlling the time of one cycle of vibrato, and the time for repeatedly reading out one waveform data in the reading means is variably controlled by the time control data. The time of the read sequence can be controlled, and the vibrato cycle can be variably controlled.

Further, corresponding to a plurality of cycles of vibrato,
By interchanging the vibrato cycle position to which each waveform data belongs while maintaining a relative time position within one vibrato cycle of the waveform data and a storage means storing a plurality of dispersive waveform data, Setting means for arbitrarily variably setting the reading order of each waveform data; and repeatedly reading out the waveform data from the storage means at predetermined intervals, and by switching the waveform data to be read out in order in the set reading order,
By providing the reading means for realizing vibrato over a plurality of cycles, the relative time position of each waveform data within one cycle of vibrato is maintained, and the vibrato cycle position to which each waveform data belongs is replaced, thereby giving a sense of incongruity. This provides an excellent effect that a vibrato sound can be synthesized in various combinations.

[Brief description of the drawings]

FIG. 1 is a flowchart showing an example of a procedure for creating a tone database according to a tone data creating method according to the present invention.

FIG. 2 is a diagram schematically showing an example of a musical score of a series of music phrases, an example of dividing a performance section corresponding to each articulation unit, and an example of analyzing musical tone elements constituting an articulation element.

FIG. 3 is a diagram showing a specific example of a plurality of tone elements analyzed from a waveform corresponding to one articulation element.

FIG. 4 is a diagram showing a configuration example of a database.

5 is an articulation sequence AE in the articulation database ADB of FIG.
SEQ and articulation element vector A
The figure which shows the specific example of EVQ.

FIG. 6 is a diagram showing a specific example of an articulation element vector AEVQ including attribute information.

FIG. 7 is a flowchart showing an example of a tone synthesis procedure according to the tone data creation method according to the present invention.

FIG. 8 is a diagram showing a configuration example of automatic performance sequence data employing a tone synthesis method according to a tone data creation method according to the present invention.

FIG. 9 is a diagram showing specific examples of some rendition style sequences according to the present invention.

FIG. 10 is a diagram showing an example of connection processing by cross-fade synthesis between articulation elements in one rendition style sequence.

FIG. 11 is a diagram outlining an example of editing a rendition style sequence (articulation element sequence).

FIG. 12 is a flowchart showing an example of an editing tree of a rendition style sequence (articulation element sequence).

FIG. 13 is a view showing a concept of a partial vector.

FIG. 14 is a flowchart partially showing a musical sound synthesis processing procedure of an articulation element including a partial vector.

FIG. 15 is a diagram illustrating an example of a vibrato synthesis process.

FIG. 16 is a diagram showing another example of the vibrato combining process.

FIG. 17 is a view showing some rules of a waveform template connection processing example;

FIG. 18 is a view showing some rules of a connection processing example of template data (envelope waveform template data) other than a waveform template. .

FIG. 19 is a view showing some concrete means of the connection rule shown in FIG. 18 (b).

FIG. 20 is a diagram showing some specific means of the connection rule shown in FIG. 18 (c).

FIG. 21 is a block diagram schematically showing a connection process of various template data and a tone synthesis process based on the template data.

FIG. 22 is a block diagram showing an example of a hardware configuration of a musical sound synthesizer according to an embodiment of the present invention.

23 is a block diagram showing a detailed example of a waveform interface in FIG. 22 and a configuration example of a waveform buffer in a RAM.

FIG. 24 is a time chart schematically showing a musical sound generation process performed based on MIDI performance data.

FIG. 25 is a time chart schematically showing a performance style performance process (articulation element musical tone synthesis process) executed based on data of a performance style sequence (articulation element sequence AESEQ).

FIG. 26 is a flowchart showing a main routine of a tone synthesis process executed by the CPU of FIG. 22;

FIG. 27 is a flowchart illustrating an example of “automatic performance processing” in FIG. 26;

FIG. 28 is a flowchart illustrating an example of “sound source processing” in FIG. 26;

FIG. 29 is a flowchart showing an example of “1 frame waveform data generation processing” for “normal performance” in FIG. 28;

30 is a flowchart showing an example of “1 frame waveform data generation processing” for “performance style performance” in FIG. 28.

FIG. 31 is a diagram conceptually showing an outline of a time axis expansion / contraction process (TSC control).

FIG. 32 is a view for explaining a hierarchical structure of a performance style sequence.

FIG. 33 is a diagram showing an example of a temporal progress state of a waveform read address when time axis compression is performed by time axis expansion / contraction control.

FIG. 34 is a diagram showing an example of a temporal progress state of a waveform read address when time axis expansion is performed by time axis expansion / contraction control.

[Explanation of symbols]

 ADB articulation database TDB template database 10 CPU 11 ROM (read only memory) 12 RAM (random access memory) 13 hard disk device 14, 15 removable disk device 16 display 17 input operation device 17 such as keyboard and mouse 18 18 waveform interface 19 Timer 20 network interface 21 MIDI interface 22 data and address bus

Claims (9)

[Claims] 1. A storage means for storing a plurality of waveform data dispersedly extracted from an original waveform to which vibrato is added, and one of the waveform data is repeatedly read out at a predetermined interval,
A vibrato generation device comprising: a read unit that executes a predetermined read sequence by sequentially switching waveform data to be read, and realizes vibrato over a plurality of cycles by repeating the read sequence. 2. The apparatus according to claim 1, further comprising control data generating means for generating control data indicating a temporal change in pitch in accordance with said read sequence, wherein said read means controls a read speed of said waveform data in accordance with said control data. The vibrato generator according to claim 1, which changes with time. 3. A control data generating means for generating control data indicating a temporal change in amplitude in response to the read sequence; and an amplitude of waveform data read by the read means, The vibrato generator according to claim 1 or 2, further comprising an amplitude control means for performing dynamic control. 4. A control data generating means for generating time control data for controlling the time of one cycle of vibrato, wherein the time for repeatedly reading out one waveform data in said reading means is variably controlled by said time control data. 4. The vibrato generation apparatus according to claim 1, wherein the vibrate period is variably controlled by controlling the time of the read sequence. 5. The image processing apparatus according to claim 1, further comprising: a cross-fade synthesizing unit that cross-fade synthesizes the preceding waveform data and the subsequent waveform data while repeatedly reading the preceding waveform data at predetermined intervals. The vibrato generator according to any one of claims 1 to 4. 6. A method of extracting a plurality of dispersed waveform data from an original waveform to which vibrato is added, storing the dispersed data in a memory, and repeatedly reading the waveform data from the memory at a predetermined interval, and reading the waveform data. Executing a predetermined read sequence by sequentially switching the waveform data, and realizing vibrato over a plurality of cycles by repeating the read sequence. 7. A storage means for storing a plurality of dispersive waveform data corresponding to a plurality of cycles of vibrato, and storing a relative time position of each of the waveform data in one cycle of the vibrato while storing each waveform data. Setting means for arbitrarily setting the reading order of each waveform data over a plurality of cycles of vibrato by replacing the vibrato cycle position to which the vibrato belongs, and waveform data to be repeatedly read out from the storage means at predetermined intervals and to be read out A vibrato generating apparatus comprising: reading means for realizing vibrato over a plurality of cycles by sequentially switching the order in the set reading order. 8. The vibrato generating apparatus according to claim 7, wherein said reading means repeats a vibrato over a plurality of cycles by repeating a read sequence according to the set read order. 9. A method of extracting a plurality of dispersed waveform data corresponding to a plurality of cycles of vibrato and storing the same in a memory, while maintaining a relative time position of each of the waveform data in one cycle of the vibrato. Arbitrarily setting the reading order of each waveform data over a plurality of cycles of vibrato by replacing the vibrato cycle position to which each waveform data belongs; and repeatedly reading and reading the waveform data from the memory at a predetermined interval. Realizing vibrato over a plurality of cycles by sequentially switching waveform data to be output in the set reading order.