JP3132392B2 - Singing sound synthesizer and singing sound generation method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a singing sound synthesizer and a singing sound generating method suitable for synthesizing chorus sounds (instrumental sounds and human voice sounds).

[0002]

2. Description of the Related Art In general, a natural musical instrument has a formant caused by a structure inherent to the musical instrument (for example, a sound board of a piano). Also, human voices have predetermined formants depending on the structure of the human body (for example, the shape of the vocal cords, vocal tract, oral cavity, etc.), which characterizes the tone color peculiar to human voices.

In an electronic musical instrument, in order to synthesize a tone closer to a natural instrument sound or a human voice sound (singing sound), voice synthesis must be performed according to a formant unique to each sound. An apparatus for performing speech synthesis using a formant as described above is disclosed in, for example, JP-A-3-200300 and JP-A-4-251297.

[0004] In order to create performance data relating to music with lyrics using such an electronic musical instrument, it is necessary to determine the sounding start timing, duration time, and the like for each of the instrumental sound and the singing sound based on the musical score. Must be set.

[0005]

However, since human voice sounds generally rise later than musical instrument sounds, there is a problem that the sounding start timing in the performance data and the sounding start timing in the auditory sense are shifted. For example, even if the sounding of the instrumental sound and the sounding of the singing sound are simultaneously started in response to the note-on signal of the instrumental sound, the singing sound sounds as if it started slightly later in terms of hearing.

In order to solve this problem, it is only necessary to adjust the timing of the performance data for the whole music or for each performance part, but the operation becomes extremely complicated. The present invention has been made in view of the above circumstances, and has as its object to provide a singing sound synthesizing apparatus and a singing sound generating method that can easily control singing sounds at appropriate timing.

[0007]

According to a first aspect of the present invention, there is provided a singing sound synthesizer for generating a predetermined singing sound corresponding to performance data. The rise time of the first phoneme generated according to the note-on signal of the performance data among the phonemes to be performed,
The first phoneme is characterized in that the rise time is shorter than the rise time when another phoneme is sounded during the note-on period.

According to a second aspect of the present invention, in the singing sound synthesizing apparatus for generating a predetermined singing sound in accordance with performance data, a rising time of a phoneme constituting the singing sound and the rising time Storage means for storing a rise characteristic within the sound data, first determination means for determining whether or not the rise time is shorter than a sounding time assigned to the phoneme; and a note-on signal of the performance data for the phoneme. A second determining means for determining whether or not the first phoneme should be pronounced in accordance with the above, and the rising characteristic is compressed in the time axis direction based on the determination results of the first and second determining means. Compression means.

According to a third aspect of the present invention, in the singing sound synthesizing apparatus for generating a predetermined singing sound in accordance with performance data, the phonemes constituting the singing sound and the duration time of each of these phonemes are provided. If a predetermined value is stored in the storage means as the duration time of the last phoneme to be lastly pronounced among the phonemes, the last phoneme is supplied until the next note-on signal is supplied. When there is an intermediate duration phoneme which is a phoneme other than the final phoneme and which has the predetermined value stored in the storage means as the duration time, a note-off signal is received, It is characterized by comprising a sounding interruption means for stopping sounding of the intermediate sustained phoneme and then sounding a phoneme after the intermediate sustained phoneme.

Further, in the configuration of claim 4, the singing sound generating a predetermined singing sound corresponding to the performance data
In the generation method, a rise time of a leading phoneme generated in accordance with a note-on signal of the performance data among phonemes constituting the singing sound is determined such that the leading phoneme is generated following another phoneme during the note-on period. The rise time is shorter than the rise time in the case of

[0011]

BEST MODE FOR CARRYING OUT THE INVENTION

1. Principle of Embodiment 1-1. Configuration of MIDI Signal Next, an electronic musical instrument according to an embodiment of the present invention will be described. First, a format of a MIDI signal supplied to the electronic musical instrument will be described with reference to FIG. Here, "C3" (do), "E3" (mi) and "G3"
It is assumed that a musical sound signal is generated by adding lyrics “sa”, “i” and “ta” to the scale of (G).

[0012] Of these MIDI signals, a portion related to a musical instrument sound will be described first. In FIG. 2, the “time” column is
The time when each MIDI signal is supplied to the MIDI interface 3 is shown. At time t ₂ , “90” “30”
The MIDI signal “42” is supplied to the MIDI interface 3. Note that in this specification, characters enclosed in double quotation marks ("" and "") represent hexadecimal numbers.

Here, "90" means note-on, "30" is the note number of "C3",
“42” is velocity. That is, MIDI signal at time t ₂ will mean "base Roshiti" 42 "case notes on the C3 sound in".

Thereafter, at time t ₃ , “90”
MIDI signals “30” and “00” are supplied. As described above, "90" means note-on, but only when the velocity is "00", exceptionally means note-off. That is, the MID at time t ₃
The I signal means "note off the sound of C3".

[0015] Similarly, at time t _5, note number "34" note-on signal ( "E3"), base Roshiti "50" is supplied at time t ₆ the note-off signal is supplied. Also, note number "37" at time t ₈ ( "G3"), base Roshiti note-on signal of "46" is supplied at time t ₉ the note-off signal is supplied.

[0016] In this way, MIDI signal of the illustrated, but within a period of time t ₂ ~t ₃ is're instructed to pronounce the sound of "C3", according to the sound of the "C3" It is necessary to specify the singing sound (lyrics) to be pronounced, ie, “sa”. In the present embodiment, such designation can be made at any time before the note-on (time t ₂ ) of the instrument sound. In the illustrated example, such instruction is made the time t ₁
And

MID first supplied at time t ₁
The I signal is “F0”. This signal is a start signal of information called “system exclusive” in the MIDI standard. The system exclusive signal “F
After the appearance of "0", this information is information until the next appearance of "F7", and the contents can be freely defined by the vendor.

Therefore, in the present embodiment, data such as singing sound is transmitted by this system exclusive. Hereinafter, such data such as singing sounds will be referred to as “phone sequence data”. By the way, system exclusive is used for various uses other than transmission of phone sequence data.

Therefore, in this embodiment, the signal “F”
Following "0", "43""1n""7F""03"
If a signal (where n is an arbitrary single digit number) is supplied, it is determined that the system exclusive is phone sequence data. Hereinafter, the signals of "43", "1n", "7F", and "03" are referred to as a phone sequence header.

Now, the signal "35" following this is
"S" means phoneme. That is, since the singing sound to be pronounced is "sa", it is decomposed into phonemes "s" and "a", and the pronunciation of "s" is instructed first. The data following the phoneme (except for the case of “00”) indicates the duration time, that is, the duration of the phoneme in units of “5 msec”.

In the illustrated example, the duration time is "0A"("10" in decimal), so that the duration time of "s" is "50 m".
sec "has been instructed. Next, the signal “20”
Indicates the phoneme “a” and “00” as the duration time.

Here, the duration time is "00".
If it is, "Continue pronunciation until the next note-on"
It means. Thus, the example shown in the figures will be at time t ₅ is note-on "E3" is sustained pronounce "a" to occur.

The reason why the undetermined duration time (until the next note-on) is designated by the signal "00" is that the lyrics are continuously produced even if the instrument sound is intermittent. This is because the tendency is strong. Of course, if lyrics should be intermittently generated, a desired value may be set as the duration time of "a" instead of "00".

Next, it is necessary to specify a singing sound to be pronounced in accordance with the sound of "E3", that is, "I". In the present embodiment, the designation is after the designated time (t ₁ ) of the immediately preceding singing sound, and is the note-on time (t ₅ ) of the musical instrument sound.
It can be done at any earlier time. In the example shown,
The time when such an instruction is performed is assumed to be t ₄ . At time t ₄ , the system exclusive start signal “F
0 ”and phone sequence header“ 43 ”“ 1n ”
“7F” and “03” are supplied.

Next, a signal "22" is supplied, which means the phoneme of "i". That is,
Since “i” is indicated by a single phoneme “i”, the pronunciation of the phoneme is instructed. For subsequent data thereto is "00", the phoneme "i" to last the sound until the next note-on i.e. time t _8, so that the instruction is made.

Next, the singing sound to be pronounced in synchronism with the sound of "G3", that is, "ta", is generated at the specified time (t ₄ ) of the immediately preceding singing sound and the note-on of the instrument sound It can be performed at any time before the time (t ₈ ). In the example shown, such instruction has been time and t _7. At time t ₇ , the system exclusive start signal and the phone sequence header are supplied again.

Next, signals "3F" and "01" are supplied. This “3F” represents the closing sound “CL” and has a meaning of “temporarily interrupt the sound”. That is, the Japanese "ta" sound does not consist solely of the "t" and "a" phonemes. The lower surface of the tongue is applied to the upper and lower incisors before the "t" phoneme is pronounced. There is a tendency for the air flow to be interrupted by contact and the sound to be interrupted once. Therefore, a closing sound "CL" of "5 msec" is provided as the first phoneme.

The signal "37" followed by "37" of the signal "02" indicates the phoneme of "t".
“20” of “0” and “00” indicates the phoneme of “a” as described above. As described in detail above, the MIDI signal supplied to the electronic musical instrument of the present embodiment is
The contents of the singing sound are specified by the sequence data, and both the instrument sound and the singing sound are instructed by the note-on signal of the subsequently generated instrument sound.

1-2. Sound Source Control Method In this embodiment, a sound source circuit similar to that disclosed in Japanese Patent Application Laid-Open No. 3-200300 is used.
In this sound source circuit, "8" channels are allocated for singing voices, and among the "4" channels, the first to fourth formants of voiced sounds are synthesized, and the remaining "4" channels are synthesized.
The first to fourth formants of the unvoiced sound are synthesized on the channel.

Here, the levels of the first to fourth formants of the unvoiced sound are UTG1 to UTG4, and these formant frequencies are UTGf1 to UTGf4. The levels of the first to fourth formants of the voiced sound are VTG1 to VTG4, and these formant frequencies are VTGf1 to VTGf4.

In the present specification, the characteristic of each phoneme in a steady state is expressed in the form of "parameter PHPAR [*]".
Here, the character "*" is replaced with a phoneme name such as "s", "a", "i". FIG. 11A shows the contents of the parameter PHPAR [*]. In the figure, the parameter PHPAR [*] includes voiced sound first to fourth formant center frequencies VF FREQ1
~ 4, unvoiced sound first to fourth formant center frequency UF FR
EQ1 ~ 4 and voiced sound 1 ~ 4 formant level VF LEV
EL1 to 4 and unvoiced first to fourth formant level UF LEV
EL1 to EL4 and formant shape designation information are included. The parameter PHPAR [*] is provided in a number corresponding to the type of phoneme (about several tens).

Next, characteristics at the time of transition from a certain phoneme to another phoneme are represented in the form of "parameter PHCOMB [1-2]".
Here, the characters "1" and "2" are "s", "a",
It is replaced by a phoneme name such as "i". For example, the parameter PHCOMB [sa] represents a characteristic when transitioning from “s” to “a”. Note that the characteristic when a certain phoneme rises from a silent state is to blank a character corresponding to “1”, such as “PHCOMB [-s]”.

Accordingly, the number of the parameters PHCOMB [1-2] is almost square times the parameter PHPAR [*]. However, the actual amount of data is much smaller than the squared number. This is because phonemes are classified into several types, such as voiced consonants, unvoiced consonants, and fricatives, and if there is a conversion characteristic of a phoneme belonging to the same classification, there is a high possibility that the same parameters can be used.

Now, the specific contents of the parameter PHCOMB [1-2] are shown in FIG. In the second row from the bottom in the figure,
A parameter called articulation coupling time COMBI TIME is provided. This parameter is a parameter indicating how much time is required for a transition of a target phoneme (for example, a transition from “s” to “a”) to be heard naturally.

Next, a parameter called a phoneme recognition time RCG TIME is provided in the lowermost column. This parameter is the time until the target phoneme starts to sound like the phoneme within the articulation combination time COMBI TIME. Therefore, the phoneme recognition time
RCG TIME is always shorter than articulation combination time COMBI TIME.

Next, the VF LEV shown in the uppermost column of FIG.
EL CURVE1 is a preceding phoneme voiced sound amplitude drop characteristic, and is a parameter indicating what characteristics the preceding phoneme, which is a voiced sound, falls within the articulation combination time COMBI TIME. Next, UF LEVEL CURVE1 is a preceding phoneme unvoiced sound amplitude decreasing characteristic, and the same content is defined for unvoiced sound.

Here, the preceding phoneme unvoiced sound amplitude drop characteristic UF LEV
As characteristics such as EL CURVE1, for example, "linear" and "exponential function" can be designated. Next, VF FREQ CURVE2
Is the following phonological voiced formant frequency change characteristics,
It specifies how to transition from the formant frequency of the voiced sound of the preceding phoneme to the formant frequency of the voiced sound of the succeeding phoneme.

UF FREQ CURVE2 is a subsequent phoneme unvoiced sound formant frequency change characteristic, and defines the same contents for unvoiced sound. Next, VF LEVEL CURVE2
Is the amplitude rise characteristic of the subsequent phoneme voiced sound, and defines the characteristics of the rise of the formant level of the voiced sound of the subsequent phoneme. Also, UF LEV
EL CURVE2 is a subsequent phoneme unvoiced sound amplitude rise characteristic, and the same content is defined for unvoiced sound.

Next, VF INIT FREQ 1 to VF INIT 4 are voiced sound first to fourth formant initial center frequencies, and are set when rising from silence (for example, parameter PHCOMB [-s]). These frequencies specify the initial values of the voiced first to fourth formant center frequencies VF FREQ1 to VF4. UF INIT FREQ1 to UF4 are unvoiced sound first to fourth formant initial center frequencies, and the same contents are defined for unvoiced sound. When rising from silence, the leading vocal voiced sound amplitude drop characteristic VF LEVEL
CURVE1 and precedence phoneme unvoiced sound amplitude drop characteristics UF LEVEL CU
RVE1 is ignored.

Here, as an example, a description will be given of a setting state in which each channel of the tone generator circuit is in the steady state of the phoneme "s" and is shifted to the steady state of the phoneme "a". First, from the timing of transition from “s” to “a”, the articulation coupling time COMBI TIME of the parameter PHCOMB [sa]
Is set as a transition period.

Then, during the period, the voiced sound first to fourth formant center frequencies of the tone generator circuit are changed based on the subsequent phoneme voiced sound formant frequency change characteristic VF FREQ CURVE2. In addition, the unvoiced sound first to fourth formant center frequencies are changed according to the subsequent phoneme unvoiced sound formant frequency change characteristic UF FREQ CURVE2.

At the same time, the voiced sounds 1 to 1 of the phoneme "s"
Fourth formant level and unvoiced sound The first to fourth formant levels are converted to the preceding phoneme voiced sound amplitude decreasing characteristics VF LEVEL C
URVE1 and subsequent phonological voiced sound amplitude rise characteristics VF LEVE
L according to CURVE2, and voiced sounds 1 to 1 of phoneme "a"
4th formant level and unvoiced sound 1st to 4th formant levels are converted to subsequent phonological voiced sound amplitude rise characteristics VF L
EVEL CURVE2 and subsequent phoneme unvoiced sound amplitude rise characteristics U
It will be raised according to F LEVEL CURVE2.

At this time, for example, the level of the first formant of the voiced sound of the sound source circuit is the sum of the level related to the phoneme “s” and the level related to the phoneme “a”. FIGS. 9 and 10 show the setting contents of each channel in accordance with the lyrics of "Sai".

In these figures, the unvoiced sound formant frequencies UTGf1 to UTGf4 and the voiced sound formant frequencies are assumed to be equal to VTGf1 to VTG4, and are expressed as "formant frequencies are TGf1 to TGf4". Further, these figures show examples of transition states of the formant level and the formant frequency, and do not show ideal transition states. 1-3. Problems of singing sound generation timing 1-3-1. Next, FIG. 12A shows the relationship between the articulation combination time COMBI TIME and the duration time when the phonemes "s" and "a" are pronounced. As described above, the articulation combination time COMBI TIME is a value uniquely determined according to the type of phoneme, and the duration time is determined by the MIDI signal.

Therefore, the value obtained by subtracting the articulation coupling time from the duration time is the sounding time (steady time) in the steady part. Here, the phoneme “s” does not sound like “s” to the human ear from the beginning of the articulation coupling time (time t _a ), but at time t _b after some time (phoneme recognition time RCG TIME) has elapsed. s ".

[0046] Therefore, in order the instrumental sound and singing sound is to play audio, as has been sounded simultaneously, as shown in FIG. (B), the timing and time of note-on of the instrument sound t _b
It is ideal to shift the sounding timing of the singing sound so that と matches.

FIG. 4B shows such an ideal state.
However, in practice, it is extremely difficult to control the pronunciation of the singing sound as shown in FIG. Because the same figure
In (b), it is necessary to start singing the singing sound before the note-on of the instrument sound, so that it is necessary to predict the note-on timing of the instrument sound that will occur in the future.

For this reason, how to match the timings of the singing sounds while setting the start timings of the singing sounds after the note-on of the musical instrument sounds is a problem. Therefore, the present inventors have studied various methods as described below.

1-3-2. Method of delaying the start timing of only the first phoneme First, the start timing of the first phoneme of the singing sound is matched with the note-on timing of the instrument sound (the start timing of the sound is delayed from the ideal form), and the subsequent phonemes are ideal form There is a method of generating sound as follows. FIG. 12C shows a transition state when such a measure is taken.

However, in this state, there is a problem that most of the stationary time of the phoneme of "s" overlaps the sounding period of the phoneme of "a", and the phoneme of "s" is hardly recognized. In other words, the listener hears a sound as if "a" is slightly noised, and it is difficult to recognize the sound as "sa".

1-3-3. Time t _b method to truncate the time t _b previous state than the previous state truncate the system ideal form of the transition state is also considered. The transition state in that case is shown in FIG. This method has a drawback that the phoneme of "s" rises rapidly and sounds very unnatural as a human voice.

1-3-4. A method of delaying the whole phoneme A method is also conceivable in which the sounding start timing of the phoneme "s" is matched with the note-on timing of the instrument sound, and the phonemes thereafter are sequentially delayed. The transition state in that case is shown in FIG.
It is shown in (b). This method has a drawback that the timing of the singing sound is delayed, so that it sounds unnatural.

1-3-5. 2. Description of the Related Art A technique of delaying a MIDI signal uniformly for a predetermined time in an electronic musical instrument to generate a sound is known (not shown). If the predetermined time is, for example, “300 msec”, the note-on timing of the instrument sound is uniformly “300 msec”.
c ".

On the other hand, for the singing sound, the delay time may be determined according to the sounding time before the note-on. For example, _if the period between the times t _{a and} t _b (phoneme recognition time RCG TIME) is “50 msec” in the ideal form, the phoneme starting from the phoneme “s” may be delayed by “250 msec”.

Thus, the phoneme can be generated in a transition state similar to the ideal form with a delay of a predetermined time. This method is optimal for reproducing a recorded MIDI signal. However, when a real-time performance is involved, the difference between the performance timing and the sounding timing becomes large, giving a very unnatural feeling to the player.

1-3-6. Method for compressing the articulation coupling time of the head phoneme According to the results of the study by the present inventors, it has been found that by compressing the articulation coupling time of the head phoneme in the time axis direction, the drawbacks of the above methods can be eliminated or alleviated. . In the above example, by compressing the transition state in the (time t _a ~t _c in FIG. 13 (a)) coarticulation time in ideal form in the time axis direction,
It will be set as the transition state from the instrument sound note-on (time t _b) to time t _c.

FIG. 13C shows a transition state when such a method is adopted. In the figure, the articulation coupling time of the phoneme "s" is short, but since it rises smoothly within that range, it is necessary to synthesize a much better voice as compared with that of FIG. 12 (d). Can be. Furthermore, since the energy of the phoneme "a" is still weak during the stationary time of the phoneme "s", the phonemes of "s" and "a" can be clearly distinguished.

2. Configuration of Embodiment Next, the configuration of the electronic musical instrument of the present embodiment will be described with reference to FIG. In the figure, reference numeral 9 denotes a CPU, which controls other components based on a program stored in the ROM 7. 8
Is a RAM, which is used as a working memory of the CPU 9. Reference numeral 1 denotes a switch panel on which conditions such as tone color are set by the user. This setting is displayed on the liquid crystal display 2.

Reference numeral 6 denotes a keyboard, which outputs performance data via the bus 10 when operated by a user.
Reference numeral 3 denotes a MIDI interface, which connects an external device to the CP.
The MIDI signal is exchanged with U9. here,
When a MIDI signal is supplied from an external device, the MIDI interface 3 interrupts the CPU 9 (MID
I reception interrupt).

Next, reference numeral 4 denotes a tone generator circuit, which generates a tone signal such as a singing sound based on the performance data supplied via the bus 10. As described above, the tone generator 4 has "4" channels for voiced and unvoiced sounds, respectively, and the CPU 9 can sequentially update the formant frequency and formant level of each channel. Reference numeral 5 denotes a sound system which generates a generated tone signal. Reference numeral 11 denotes a timer, which generates an interrupt (timer interrupt) to the CPU 9 every predetermined time.

3. Operation of embodiment 3-1. Initial Operation First, when the electronic musical instrument of the present embodiment is started, a main routine shown in FIG. 3 is started. In the figure, when the process proceeds to step SP1, predetermined initial settings are performed. Next, when the process proceeds to step SP2, task management is performed. That is, a plurality of routines (tasks) are switched and executed in parallel according to the interrupt state.

Of these, the routine with the highest priority is
MID activated when MIDI reception interrupt occurs
This is an I reception interrupt processing routine. The next highest priority routine is a timer interrupt processing routine started by a timer interrupt.

The other routines have the lowest priority. A performance data processing routine, which will be described later, is provided as one of the low-priority routines, and is operable when the interrupt processing routine is in a pause state.

3-2. Reception of Phone Sequence Data (Time t ₁ ) When a MIDI signal is supplied at the MIDI interface 3 or an event occurs at the keyboard 6, a MIDI reception interrupt processing routine shown in FIG. 4 is started. In the figure, the processing is step SP1
At step 1, the contents of the MIDI signal or the keyboard 6
Is written in a predetermined area (MIDI reception buffer) in the RAM 8, and the process is immediately terminated.

Here, the operation information of the keyboard 6 is note-on information with note number and velocity, note-off information with note number, etc., and has the same contents as the MIDI signal related to the instrument sound. . Therefore,
In this specification, a MIDI signal supplied from the MIDI interface 3 and operation information output from the keyboard 6 are collectively referred to as a “MIDI signal”. Hereinafter, at time t ₁
In ~t _9, sequential M is MIDI signal shown in FIG. 2
The operation of the present embodiment will be described on the assumption that the data is stored in the MIDI reception buffer via the IDI interface 3.

At time t ₁ , when the phone sequence data relating to the sound of “sa” is stored in the MIDI reception buffer, the performance data processing routine is started at an appropriate time thereafter (when no interruption occurs). . In FIG. 5, when the process proceeds to step SP21, a MIDI signal of "1" bytes is fetched from the MIDI reception buffer.

In the example of FIG. 2, the MI supplied at time t ₁
Since the first byte of the DI signal is "F0", this "F0" is taken in. Next, when the process proceeds to step SP22, it is determined whether or not the received MIDI signal is a "status byte" (a value of "80" to "FF"). Here, since the determination is "YES", the process proceeds to step SP24, and the type of the status byte (in this case, a system exclusive start signal) is stored in a predetermined area of the RAM 8.

Next, when the processing proceeds to step SP25,
Processing branches depending on the type of the status byte. If the status byte is system exclusive, the process proceeds to step SP27. Here, a “4 byte” MIDI signal following the system exclusive start signal is read from the MIDI reception buffer,
It is determined whether the content is a phone sequence header.

In the example shown in FIG. 2, “F” at time t ₁
"43", "1n", "7F", and "03" following "0" are read in. Since this is nothing but a phone sequence header, in step SP27, "YE" is read.
S ", and the process proceeds to Step SP28.

In step SP28, the MID is kept until the system exclusive end signal "F7" appears.
The contents of the I reception buffer are repeatedly read, and the phone sequence data stored therein is read, and the contents are stored in a predetermined area phoneSEQbuffer in the RAM 8. In the example shown, the phonemes "s", "a" and their duration times are in the region phoneSEQbuffe
will be stored in r.

Further, in step SP28, the number of phonemes (here, "2") is substituted for the variable phone number. When the above process ends, the process of this routine ends once. Thereafter, when a timer interrupt occurs every “5 msec”, a timer interrupt processing routine shown in FIG. 8 is started.

In the figure, when the processing proceeds to step SP61, it is determined whether or not there is a phoneme being sounded.
As soon as it is determined to be "O", the process ends. In the above example, the phone sequence captured at time t ₁
None of the phonemes contained in the data are pronounced,
No substantial processing is performed in the timer interrupt processing routine.

3-3. The note-on of “C3” is thereafter time t ₂ , and the note-on of “C3” is M
When supplied to the IDI interface 3, the note-on is written to the MIDI reception buffer by executing the MIDI reception interrupt processing routine. Then, the performance data processing routine is started again.

[0074] The processing in FIG. 5 proceeds to step SP21, the first byte "90" of the received MIDI signal is captured at time t _2. Since this is a status byte, the processing proceeds to step S22 through step SP22.
Proceed to P24.

When the first byte of the MIDI signal is "90", it is either note-on or note-off. Therefore, in step SP24, if the leading byte is "90", the data of the following "2" bytes is read out, and it is determined whether the data is note-on or note-off.

In the above example, the data following "90" is "30" and "42", and the velocity ("42") is a value other than "00". It is determined that there is, and the result is stored in the RAM 8. Next, in view of the content of this determination, the process proceeds to step SP31 (FIG. 6) via step SP25.

In step SP31, the variable phone
SEQtime_counter and variable phoneSEQphone_counter
Are both substituted with “0”. Where the variable phoneSEQpho
ne_counter is the phoneme ("s") in this note
And "a") are variables that indicate the phonemes that are currently being pronounced.

That is, the number of the first phoneme is set to “0”, and the number incremented by “1” corresponding to the subsequent phonemes becomes the content of the variable phoneSEQphone_counter. Also, the variable phoneSEQtime_counter is a variable for counting the elapsed time after the start of the sounding of the phoneme to be currently sounded in units of “5 msec”.

Next, when the processing proceeds to step SP32,
It is determined whether or not data called “expiration information” exists in the head “1” note (phone sequence data supplied at time t ₁ in the above example) in the area phoneSEQbuffer. Here, the exhalation information is exhalation (breath)
Is performed, and a predetermined number is assigned similarly to other phonemes.

Here, since no breath information exists, "N
It is determined as "O", the process proceeds to step SP33, and the expiration flag f_koki is set to "0". Next, when the process proceeds to step SP35, from the area phoneSEQbuffer,
The leading phoneme number and its duration time are extracted.

In the above example, the phoneme number “35” of “s” and the duration time “0”
A "is extracted. Next, the processing proceeds to step S
In P36, the parameter PHPAR [*] and the parameter PHCOMB [1-2] are read from the database in the ROM 7 based on the preceding and succeeding phoneme relationships. In this example,
Since the phoneme "s" rises from silence, the parameter PHPA
R [s] and the parameter PHCOMB [-s] will be read.

Next, when the process proceeds to step SP37,
Articulation time COMBI TIME in parameter PHCOMB [-s]
Is determined to be less than or equal to the duration time of the phoneme “s”. Here, if "NO" is determined, step SP38 is executed, and the articulation coupling time is reset to the value of the duration time.

In any case, the processing then proceeds to step SP3
Proceeding to 9, the change characteristic applied to the pronunciation of the current phoneme (s) is calculated. However, as a prerequisite, it is necessary to compress the articulation coupling time, or the step SP
If compressed at 38, the articulated coupling time after compression is applied.

In the above example, the phoneme "s" is
Since it is located immediately after the sequence header, it is a phoneme that is pronounced with the note-on of the instrument sound. Therefore, FIG.
Each change characteristic stored in the database is compressed in the time axis direction according to the rules described in 3 (a) and (c).

That is, these change characteristics are originally change characteristics within the normal articulation coupling time COMBI TIME, but these change characteristics are changed so that the transition is completed within the time “COMBI TIME−RCG TIME”. It is compressed in the time axis direction. Further, even if the phoneme is pronounced after “s”, if step SP38 is executed first, these change characteristics are compressed based on the new (shortened) articulation coupling time. is there.

Next, each change characteristic (property appropriately compressed)
In, formant data corresponding to the current value of the variable phoneSEQtime_counter (here, “0”) is calculated. Next, when the process proceeds to step SP40, the calculated formant data is written to the singing sound channel of the tone generator circuit 4.

In step SP 40, when the singing sound channel is in the note-off state, these note-on signals are also supplied to the tone generator 4. In the above example, since the phoneme “s” is the first singing sound in the music, the note-on signal is also supplied to the sound source circuit 4.

Thus, the singing sound of the phoneme “s” is started to be generated. Further, step SP40
, The note-on signal of the instrument sound itself
It goes without saying that it is supplied to When the above processing ends, the performance data processing relating to the current note-on ends.

3-4. Timer Interrupt During Sound Generation Next, when a timer interrupt occurs, the timer interrupt processing routine of FIG. 8 is started. In this case, since the phoneme “s” is being generated, “YE” is determined in step SP61.
S "is determined, and the process proceeds to step SP62.

Here, it is determined whether or not the variable phone_duration_time is "0" (= "00"), that is, whether or not the duration time is undetermined. The duration time of the phoneme “s” is “10” (= “0”).
A "), the determination is" NO "here, and the process proceeds to step SP63. Here, the variable phoneSEQt
It is determined whether or not ime_counter has exceeded the duration time.

In the above example, the variable phoneSEQtime_co
unter is previously set to "0" in step SP31. On the other hand, the duration time of the phoneme “s” is “1”.
0 "(=" 0A "). Therefore, here, "YE
S ”, the process proceeds to step SP64, and the variable p
honeSEQtime_counter is incremented by "1".

Next, when the process proceeds to step SP65,
In each of the change characteristics (compressed) calculated in step SP39, the current variable phoneSEQtime_counter
(Here, “1”) is calculated.

Next, when the process proceeds to step SP66,
The calculated formant data is written to the singing sound channel of the tone generator 4. This gives the phoneme "s"
Is in a state advanced by “5 msec” in each of the above change characteristics. From the above, 1
The timer interrupt processing for the current cycle ends.

Thereafter, the routine is started every "5 msec", and in step SP64, the variable phoneSEQtime_
The counter is incremented by "1". Then, based on the incremented value, step SP
Steps 65 and 66 are executed.

Thus, the formant data of the tone generator 4 is updated so that the phoneme "s" gradually rises. If the duration time is longer than the articulation combination time, the steady-state phoneme "s" is generated based on the parameter PHPAR [s] for a time corresponding to the difference.

3-5. Transition to phoneme "a" By the way, if the increment processing of step SP64 is repeated, the variable phoneSEQtime_counter will eventually exceed the duration time. Thereafter, the timer interrupt processing routine is called again, and the process proceeds to step S.
When the process proceeds to P63, “NO” is determined here, and the process proceeds to step SP67.

Here, the variable phoneSEQphone_counter
Is incremented by “1” to become “1”. That is, the variable indicates “a” which is the “second” phoneme. And correspondingly, the variable phoneSE
Qtime_counter is reset.

Next, when the process proceeds to step SP68,
It is determined whether the variable phoneSEQphone_counter is less than or equal to the variable phone number. Since "2" has been substituted for the variable phone number in step SP28, "YES" is determined here, and the process proceeds to step SP28.
Go to 69.

Here, from the region phoneSEQbuffer,
The “second” phoneme number and its duration time are extracted. In the above example, the phoneme number “a” of “a” and its duration time “00” are extracted.

Next, when the processing proceeds to step SP70,
The parameter PHPAR [*] and the parameter PHCOMB [1-2] are read from the database in the ROM 7 based on the preceding and succeeding phoneme relationships. In this example, since the state of the sound source circuit 4 is to be changed from the phoneme “s” to the phoneme “a”, the parameter PHPAR [a] and the parameter PHCOMB [sa] are read.

Next, when the process proceeds to step SP65,
For each change characteristic included in this parameter PHCOMB [sa], formant data corresponding to the value of the current variable phoneSEQtime_counter (here, “0”) is calculated. Next, when the process proceeds to step SP66, the calculated formant data is written to the singing sound channel of the tone generator circuit 4. As a result, the transition from the phoneme “s” to “a” is started.

Thereafter, as in the operation described for the phoneme "s", the routine is started every "5 msec", and the variable phoneSEQtime_counter is incremented by "1" at step SP64. Then, based on the incremented value, steps SP65 and SP66
Is executed.

As a result, the formant data of the tone generator 4 is updated so that the phoneme “s” gradually transitions to the phoneme “a”. Eventually, the parameter PHCOMB [s-
When the articulation combination time COMBI TIME relating to a] elapses, the phoneme “a” in the steady state is sounded. In this case, since the duration time is "0", step SP63 is always skipped.

3-6. Note-off of “C3” At time t ₃ in FIG. 2, when the MIDI signal of “C3” note-off is supplied to the MIDI interface 3,
The contents are written to the MIDI reception buffer by the MIDI reception interrupt processing routine (FIG. 4).

Thereafter, a performance data processing routine (FIG. 5)
Is activated, the note-off signal is captured in step SP21, and the process proceeds to step SP51 (FIG. 7) via steps SP22 to SP25. here,
After the “phoneme whose duration time is 0”, it is determined whether or not another phoneme exists.

[0106] The term "phoneme duration time is 0" here is a phoneme "a", in the notes that have been supplied to the time t _1, other phonemes after this phoneme "a" does not exist. Therefore, the determination is "NO" here, and the process proceeds to step SP57.

Here, it is determined whether or not the expiration flag f_koki is “1”. Since the expiration flag f_koki is set to “0” in step SP33,
Here, the determination is “NO”, and the process proceeds to step SP59.
To perform key-off processing of the instrument sound.

Thus, the performance data processing relating to the note-off processing ends. That is, in this example, it is understood that processing that directly affects the singing sound is not performed in response to the note-off of the instrument sound. Therefore, even after the note-off,
The pronunciation for the phoneme "a" is maintained.

3-7. Subsequent Note-On Next, at time t ₄ , when the phone sequence data relating to the phoneme “i” is supplied to the MIDI interface 3, the data is written to the MIDI reception buffer in the MIDI reception interrupt processing routine. Thereafter, in step SP28 of the performance data processing routine, the phone sequence data is written into the buffer phoneSEQbuffer, and "1" is substituted for the variable phone number.

[0110] Next, when the note-on signal of the instrument sound "E3" is supplied at time t _5, so that once again note-on process shown in FIG. 6 is executed. At that time, step S
The parameters referred to in P36 are PHPAR [i] and PHCOMB [a
-i].

Since the phoneme “i” is a phoneme that is generated in accordance with the note-on signal, PHCOMB [ai] is used in step SP39 in the same manner as when the phoneme “s” rises.
The articulation coupling time COMBI TIME is shortened, and accordingly, each change characteristic is also compressed in the time axis direction.

As a result, the singing sound transitions from "a" to "i", and the singing sound enters the steady state of "i". Thereafter, when a new note-on signal relating to the instrumental sound is generated, the phoneme number and the duration time of the next phoneme are read out, and the singing sound sequentially transitions.

3-8. Processing Corresponding to Expiration Information By the way, the phone sequence data can include various information other than those described above. One of them is breath information (breathing). Therefore, a process in the case where the exhalation information is included in the phone sequence data will be described.

After the phone sequence data including the exhalation information is supplied, when the note-on of the corresponding musical instrument sound occurs, the processing shown in FIG. 6 is executed as described above.
Then, in step SP32, "YES" is determined, and the expiration flag f_koki is set to "1".

Thereafter, the same processing as when no exhalation information is included is performed.
Is performed, "YES" is determined in the step SP57, and a singing sound key-off process is performed in a step SP58.

That is, a key-off signal of the singing sound is supplied to the sound source circuit 4. Then, the sound source circuit 4 performs a release process of gently dropping the singing sound. As a result, there is a silent state between the currently processed note and the subsequent note-on, and the singing sound is produced as if a human were breathing. 3-9. The case where the duration time is finite The processing when the duration time of all phonemes included in one note is a finite value (a value other than “00”) will be described. In such a case, each time the timer interrupt processing routine (FIG. 8) is started, the variable p
honeSEQtime_counter is incremented, and the next time the routine is started, the value of the variable is compared with the duration time of the phoneme being processed.

Then, in step SP63, "YE
If "S" is determined, step SP67 is executed,
The variable phoneSEQphone_counter is incremented. When the duration time for the final phoneme has elapsed, the variables phoneSEQphone_counter and ph
Since one number is equal, “NO” is determined in the step SP68, and the process proceeds to a step SP71.

In step SP71, a key-off process is performed on the singing sound. That is, the key-off signal of the singing sound is supplied to the tone generator circuit 4, and a silent state is provided between the currently processed note and the subsequent note-on. Such a finite duration time is suitable for use in staccato, that is, in the case of intermittently producing a singing sound.

3-10. Duration time is "0"
The case where another phoneme follows the phoneme whose duration is "0" means that the phoneme included in one note is in the order of "s", "a" and "t". In this case, the duration time of “a” is “00”, and the duration times of “s” and “t” are finite values.

In such a case, when note-off of the corresponding instrument sound occurs and the processing proceeds to the routine of FIG. 7, step SP
51 is determined as “YES”, and the process proceeds to step S
Proceed to P52. Here, the variable phoneSEQphone_counter
Is set to a value indicating the phoneme immediately after the phoneme whose duration time is “0”.

In the above example (“s”, “a”, “t”), the variable phoneSEQphone_counter is set to “2” indicating “t”. Further, in step SP52, the variable phoneSEQtime_counter is set to "0".

Next, when the processing proceeds to step SP53,
The next phoneme number and duration time are extracted from the buffer phoneSEQbuffer. That is, in the above example, the phoneme number of “t” and its duration time are extracted.

Next, when the process proceeds to step SP54,
The parameter PHPAR [*] and the parameter PHCOMB [1-2] are read from the database in the ROM 7 based on the preceding and succeeding phoneme relationships. In this example, the parameter PHPAR
[t] and the parameter PHCOMB [at] are read.

Next, when the process proceeds to step SP55,
For each change characteristic included in these parameters, formant data corresponding to the current value of the phoneSEQtime_counter variable (here, “0”) is calculated. Next, when the process proceeds to step SP56, the calculated formant data is written to the singing sound channel of the tone generator circuit 4. As a result, the singing sound starts to transition from the phoneme “a” to “t”.

Next, when the process proceeds to step SP59,
Key-off processing of the instrument sound is performed. Thereafter, the timer interrupt processing routine (FIG. 8) is repeatedly executed, so that the singing sound transitions from “a” to “t”, and a sound is generated in the steady state of “t”.

When the duration time of the final phoneme “t” has elapsed, the variable phoneSEQphone_coun
Since ter is equal to the variable phone number, "NO" is determined in the step SP68. Thus, in step SP71, a key-off process for the singing sound is performed.

As described above, a phoneme (“a”) whose duration time is “0” is followed by another phoneme (“t”).
Continues, the latter sounding is started at the timing when the note-off of the instrument sound occurs. Therefore, except for special situations (for example, when the duration time of the “other phoneme” is extremely long or when the timing until the note-on of the next instrument sound is extremely short), the note-on of the next instrument sound is performed. All phonemes can be completed before the sound is generated.

4. Modifications The present invention is not limited to the embodiments described above,
For example, various modifications are possible as follows. 4-1. In the above-described embodiment, when the exhalation information is included in the phone sequence data, the key-off process of the singing sound is performed at the time of the note-off of the instrument sound (FIG. 7, steps SP57 and SP58). However, before the key-off process, a breathing sound (a sound of a human being breathing) may be generated.

4-2. In the above embodiment,
The sound source circuit 4 outputs “4” for each of voiced sound and unvoiced sound.
Each channel was reserved, but for phonemes with strong high-frequency components like "s", additional channels were reserved.
A formant for a high frequency component may be generated. FIG.
FIG. 10 shows the frequencies and formant levels of such additional formants as “TGf5” and “UTG5”.

4-3. In the above embodiment,
Although the articulation coupling time COMBI TIME uses a common value for all formants, a different value may be used for each formant. In addition, the timing at which each formant starts transition may be different from each other.

4-4. In the above embodiment,
FIG. 13 shows a specific example of shortening the rise time of an audio signal.
Although the technique of changing the formant level is used as shown in (c), various methods for shortening the rise time can be used instead.

4-5. In the above embodiment,
Although the ROM 7 is used as the storage medium for storing the program, it goes without saying that the present invention can be realized on a storage medium such as a CD-ROM or a floppy disk as software for a personal computer. Further, the present invention can be realized on software including the sound source circuit 4, and can be applied not only to electronic musical instruments but also to various amusement devices such as game machines and karaoke machines.

[0133]

As described above, according to the first and fourth aspects of the present invention, the rise time of an audio signal generated with a note-on signal is made shorter than the rise time of another audio signal. Can be. Further, according to the configuration of the second aspect, even when the rising time is shorter than the sounding time assigned to the audio signal, the rising characteristic can be compressed in the time axis direction. Thus, an audio signal can be generated at an appropriate timing with note-on.

Further, according to the configuration of the third aspect, since the sounding of the intermediate continuous phoneme can be stopped in response to the reception of the note-off signal, the subsequent phonemes can be sounded at appropriate timing.

[Brief description of the drawings]

FIG. 1 is a block diagram of an electronic musical instrument according to an embodiment of the present invention.

FIG. 2 is a diagram illustrating an example of a MIDI signal supplied to an embodiment;

FIG. 3 is a flowchart of a main routine according to one embodiment.

FIG. 4 is a flowchart of a MIDI reception interrupt processing routine according to one embodiment;

FIG. 5 is a flowchart of a performance data processing routine according to one embodiment.

FIG. 6 is a flowchart of note-on processing of a performance data processing routine.

FIG. 7 is a flowchart of a note-off process of a performance data processing routine.

FIG. 8 is a flowchart of a timer interrupt processing routine according to one embodiment.

FIG. 9 is a diagram illustrating an example of a formant frequency and a formant level set for each channel of the sound source circuit 4;

FIG. 10 is a continuation of FIG. 9;

FIG. 11 is a diagram illustrating a data format of parameters stored in a database according to an embodiment.

FIG. 12 is a diagram illustrating various examples of a phoneme transition method when a note-on event occurs.

FIG. 13 is a continuation of FIG. 12;

[Explanation of symbols]

1. Switch panel 2. Liquid crystal display 3.
... MIDI interface, 4 ... tone generator, 5 ...
Sound system, 6 Keyboard, 7 ROM
(Storage medium), 8 RAM (storage means), 9 CP
U (first and second determination means, compression means, sounding continuation means, sounding interruption means), 10 ... bus, 11 ... timer.

Claims (4)

(57) [Claims] 1. A singing sound synthesizing apparatus for generating a predetermined singing sound in accordance with performance data, wherein a rising edge of a leading phoneme generated in accordance with a note-on signal of said performance data among phonemes constituting said singing sound. A singing sound synthesizer characterized in that the time is shorter than the rise time when the first phoneme is generated following the other phonemes during the note-on period. 2. A singing sound synthesizing apparatus for generating a predetermined singing sound in accordance with performance data, comprising: storage means for storing a rise time of a phoneme constituting the singing sound and a rise characteristic within the rise time. First determining means for determining whether or not the rise time is shorter than a sounding time assigned to the phoneme; and the phoneme is a first phoneme to be sounded with a note-on signal of the performance data. A second judging means for judging whether or not the rising characteristic is reduced in a time axis direction based on the judgment results of the first and second judging means. Singing sound synthesizer. 3. A singing sound synthesizing apparatus which generates a predetermined singing sound in accordance with performance data, wherein: a storage means for storing phonemes constituting said singing sounds and duration times of these phonemes; When a predetermined value is stored in the storage unit as the duration time of the last phoneme to be last pronounced, the sounding continuation unit for continuing the sounding of the final phoneme until a next note-on signal is supplied; If there is an intermediate continuous phoneme other than the final phoneme and in which the predetermined value is stored in the storage unit as the duration time, the sound generation of the intermediate continuous phoneme is stopped in response to the reception of the note-off signal. A singing sound synthesizing device, further comprising: a pronunciation suspending unit for producing a phoneme after the intermediate continuous phoneme. 4. A singing sound generating method for generating a predetermined singing sound in accordance with performance data, wherein a phoneme constituting the singing sound includes a first phoneme generated along with a note-on signal of the performance data. A method for generating a singing sound, characterized in that the rise time is shorter than the rise time when the first phoneme is generated following another phoneme during the note-on period .