JPH1031496A - Musical sound generating device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a musical sound generating device which can generates a musical sound so that synthesized human voice is like a natural singing voice. SOLUTION: A CPU 4 sets musical sound information consisting of note-ON and note-OFF information a pitch frequency, and a velocity in a DSP 9 according to externally supplied MIDI data, and reads voice synthesis parameters of a syllable corresponding to a text (one word) made to correspond to a musical sound to be noted ON out of a data base in a data ROM 8 and sets them in a RAM 10. Oonsequently, the DSP 9 vocalizes a human voice, which is subjected to PARCOR(partial auto-correlation) synthesis according to given speech synthesis parameters, as natural singing according to the musical sound information.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice synthesizing device, and more particularly to a tone generating device capable of generating a natural singing voice.

[0002]

2. Description of the Related Art Conventionally, as a method of synthesizing a human voice based on feature parameters extracted by voice analysis, channel vocoder, linear prediction, PARCOR (Percall)
A technique called is known. These speech synthesis techniques can convert the analyzed speech into a smaller amount of information,
In other words, it focuses on analyzing speech, converting it into the form of feature parameters, and compressing the amount of information excluding redundant components not related to the meaning of words. He did not consider applying the human voice to the formation of musical sounds.

In such a situation, the channel vocoder has a simple structure and is suitable for real-time analysis and synthesis. Therefore, the channel vocoder has been applied to a tone generator which synthesizes a tone based on the power spectrum envelope of speech extracted by a filter bank. . However, in the channel vocoder, high-quality sound synthesis was not achieved due to the limitation of the number of band-pass filter stages constituting the filter bank and problems such as inability to synthesize consonants.

[0004]

On the other hand, in the conventional tone generator using the waveform memory reading method, the simplest method is to store the sampled human voice in the waveform memory and read and reproduce it at the sampling pitch. Although it is possible to generate a high-quality human voice in a simple form, if you try to read and play it at a pitch different from the pitch at the time of sampling, the formant frequency of the human voice will change according to the conversion pitch amount However, there is a problem that a natural singing voice cannot be generated.

SUMMARY OF THE INVENTION It is an object of the present invention to provide a musical sound generator capable of forming a human voice sound synthesized as a natural singing voice.

[0006]

In order to achieve the above object, according to the first aspect of the present invention, parameter storage means for storing speech synthesis parameters for each of a plurality of syllables, musical tone information representing each musical tone of a music piece, Music information generating means for generating utterance information representing lyrics associated with each musical tone;
Parameter generating means for reading out and outputting speech synthesis parameters of syllables matching the lyrics from the parameter storage means in accordance with the utterance information generated by the music information generating means; and a person synthesizing based on the speech synthesis parameters output by the parameter generating means. Voice synthesizing means for generating a vocal sound as a singing voice according to the musical tone information.

According to the second aspect of the present invention, the parameter storage means includes an identifier indicating the end of data in a speech synthesis parameter for each syllable, and the speech synthesis means comprises: When this identifier is detected and it is determined that the voice synthesis parameter has reached the end of the data, the singing voice being synthesized is continuously pronounced.

[0008] Similarly, claim 3 is dependent on claim 1.
According to the invention described in (1), the voice synthesizer variably controls the sound quality of the synthesized human voice in accordance with the velocity included in the musical sound information.

According to the fourth aspect of the present invention, the parameter generating means expresses the utterance information by an exclusive message of MIDI data.

[0010] Further, claim 5 is dependent on claim 1 above.
According to the invention described in (1), the voice synthesizing means controls the pronunciation and mute of the singing voice by note-on note-off included in the musical tone information.

According to the present invention, the parameter storing means stores voice synthesis parameters for each of a plurality of syllables, and the music information generating means generates musical sound information representing each musical sound of the musical composition and utterance representing the lyrics associated with each musical sound. When information is generated,
The parameter generating means reads out voice synthesis parameters of syllables matching the lyrics from the parameter storage means according to the utterance information. Then, the voice synthesizing unit synthesizes a human voice based on the voice synthesizing parameter output from the parameter generating unit, and sounds this as a singing voice according to the musical tone information. As a result, it is possible to form a musical tone as a natural singing voice using the synthesized voice.

[0012]

BEST MODE FOR CARRYING OUT THE INVENTION The musical sound generating apparatus according to the present invention can be applied to an electronic musical instrument as well as an apparatus for providing voice guidance using human voices. Hereinafter, a tone generator according to an embodiment of the present invention will be described as an example with reference to the drawings.

A. 1. Configuration of Embodiment (1) Overall Configuration FIG. 1 is a block diagram showing the overall configuration of a musical sound generating device according to an embodiment of the present invention. In FIG. 1, reference numeral 1 denotes a keyboard section, which generates performance information including a key-on / key-off signal, a key code, and a velocity signal according to a key press / release operation. Reference numeral 2 denotes a panel switch group provided on the operation panel surface. The panel switch group 2 includes a power switch for turning on and off the power, a mode switch for selecting an operation mode, and the like, and generates an operation signal in accordance with the operation of these switches.

The operation mode referred to here is a keyboard mode in which a synthesized voice is produced as a singing voice according to performance information output from the keyboard unit 1 in response to a key press or release operation, or MIDI. MID supplied from an external MIDI device via an interface circuit 7 (described later)
This refers to a MIDI mode in which a singing voice is generated according to the I data. In this embodiment, the MIDI mode will be described.

Next, reference numeral 3 denotes a display unit composed of a liquid crystal panel or the like, which displays a setting state and an operation state corresponding to the operation of the panel switch group 2. Reference numeral 4 denotes a CPU for controlling each unit of the apparatus, and its characteristic operation will be described later in detail. The CPU 4 transfers the speech synthesis program stored in the ROM 5 to the DSP 9 (described later),
A voice synthesis parameter corresponding to MIDI data supplied from the outside via the MIDI interface circuit 7 is read from a data ROM 8 described later and transferred to the RAM 10.

The ROM 5 stores various control programs and control data loaded into the CPU 4 and a DS (described later).
A speech synthesis program executed in P9 is stored. The ROM 5 also has a conversion table TBL for converting note numbers in MIDI data to pitch frequencies.
Is stored. As shown in FIG. 2, this conversion table TBL generates a corresponding pitch frequency PF using a note number as a read address.

Reference numeral 6 denotes a RAM for temporarily storing various registers and flag data as a work area of the CPU 4.
The RAM 6 has a buffer area for temporarily storing serial MIDI data supplied from the outside via the MIDI interface circuit 7. The MIDI interface circuit 7 fetches MIDI data supplied from an external MIDI device under the instruction of the CPU 4, and receives the MIDI data via the bus B.
Store in the buffer area of AM6. Also, this MI
The DI interface circuit 7 sends out the MIDI data stored in the buffer area of the RAM 6 in a serial format.

In the data ROM 8, speech synthesis parameters for each syllable are registered in a database. The speech synthesis parameter is composed of a residual Z obtained by a Parcol analysis (PARCOR) analysis and Percoll coefficients K1 to Kn. In the data ROM 8, speech synthesis parameters for each syllable, that is, for each kana character, are stored in a table. I have.

Here, the data structure of the speech synthesis parameters stored in the data ROM 8 will be described with reference to FIG. In this figure, DB [m, n] is a header representing a tone number m and a coefficient set number n. This header DB
[M, n] can be used to search for the residual Z of the desired syllable and the Percoll coefficients K1 to Kn. The sound number m of the header DB [m, n] specifies a syllable (one word). On the other hand, the number of coefficient sets n is the number of Percoll coefficients K1 to Kn used when synthesizing the syllable specified by the sound number m.
And the residual Z.

For example, DB [0,0] to DB shown in FIG.
[0, n-1] is a time series of values of the residual Z and the values of the Percoll coefficients K1 to Kn obtained by Percoll analysis of "syllable" A "" specified by the note number 0.
Normally, the coefficient set number n is a value obtained by Percoll analysis at a time interval of about 5 to 20 msec. At this time interval, the Percoll coefficients K1 to Kn and the residual Z specified by the coefficient set number n are applied to a synthesis filter (described later). If set, the same formant sound as when Percoll analysis was performed is synthesized. The feature of the speech synthesis parameters compiled in the database is that DB [0 to m,
n] is provided with an identifier indicating the end of data with the residual Z set to “−1”, and the intended purpose will be described later.

Next, returning to FIG. 1, the configuration of the embodiment will be described. In FIG. 1, a DSP 9 is a CPU
4 is to generate a human voice by performing a Percoll synthesis operation based on a voice synthesis program and a voice synthesis parameter transferred and set from Step 4. The Percoll synthesis algorithm will be described later. Reference numeral 10 denotes a RAM 10 used as a work area of the DSP 9, and a speech synthesis parameter (residual Z and Percoll coefficients K1 to K1) read from the data ROM 8 by the CPU 4.
n) is set.

The audio data synthesized by the DSP 9 is converted into an analog audio signal via the D / A converter 11 at the next stage. The audio signal output from the D / A converter 11 is subjected to filtering such as unnecessary noise removal by the amplifier 12 and then amplified, and is emitted as a natural singing voice from the speaker SP.

(2) Parcall synthesis algorithm (functional configuration) of DSP 9 Next, with reference to FIG. 4, a percall synthesis algorithm (functional configuration) of the DSP 9 that forms a singing voice based on a human voice according to the musical tone information supplied from the CPU 4 will be described. I do.
Note that the tone information referred to here is a note-on NON and a note-off NOF extracted from the MIDI data by the CPU 4.
And the pitch frequency P obtained by converting the note number in addition to the velocity data VEL using the conversion table TBL.
F.

In FIG. 4, reference numeral 20 denotes a pulse generator which outputs a pulse waveform PW at a period corresponding to the pitch frequency PF. Reference numeral 21 denotes a noise generator that generates white noise WN. SEL is a selector for selecting the pulse waveform PW or the white noise WN depending on whether the voice to be synthesized is “voiced sound” or “unvoiced sound”. That is, the selector SEL selects the pulse waveform PW side to synthesize the "voiced sound" when the given Percoll coefficient K1 is equal to or larger than the predetermined constant J (about 0.3), while the Percoll coefficient K1 is When it is smaller than the constant J, the white noise WN side is selected to synthesize "unvoiced sound".

Reference numeral 22 denotes an envelope generator.
An envelope waveform ENV whose attack part, release part, and amplitude are controlled in accordance with note-on NON, note-off NOF, and velocity data VEL is generated. 2
A coefficient multiplier 3 multiplies either the pulse waveform PW or the white noise WN supplied via the selector SEL by the envelope waveform ENV and outputs the result.

By the way, in the selector SEL described above,
Switching to the pulse waveform PW or the white noise WN according to the voiced / unvoiced sound is not limited to this, and the pulse waveform PW and the white noise WN may be cross-fade according to the voiced / unvoiced sound. Then, the change from voiced sound to unvoiced sound or the change from unvoiced sound to voiced sound becomes more natural.

A coefficient multiplier 24 multiplies the output of the coefficient multiplier 23 by a residual Z and outputs the result. 25-1 to 2
Reference numeral 5-n denotes a lattice filter that synthesizes speech in the reverse process of the Percoll analysis process based on the Percoll coefficients K1 to Kn, respectively. These cascaded grating filters are:
It simulates the vocal tract characteristics, and includes a delay circuit 25a,
It comprises coefficient multipliers 25b and 25c, adder 25d and subtractor 25e. The delay circuit 25a has the same formant as the analyzed voice signal if the sampling delay is the same as in the Percoll analysis process. Therefore, when the formant is intentionally made different as a special effect at the time of speech synthesis, the sampling delay amount may be different from that at the time of Percoll analysis.

According to such a Percoll synthesis algorithm, the speech synthesis parameters (residual Z and Percoll coefficients K1 to Kn) stored in the RAM 10, that is,
The CPU 4 responds to the MIDI data by
A human voice generated by synthesizing syllables according to the voice synthesis parameters read from M8 is generated as a singing voice.

B. Next, the operation of the embodiment having the above configuration will be described with reference to FIGS. Hereinafter, the processing of the main routine will be described first as an overall operation, and then the processing of the MIDI processing routine and the timer interrupt routine called in the main routine will be described.

(1) Operation of Main Routine When the power of the tone generator according to the present embodiment is turned on, the CPU 4 reads a predetermined control program from the ROM 5 and loads it into itself, and then executes the main routine shown in FIG. Is executed, and the process proceeds to step SA1 to initialize various register flags provided in the RAM 6 and the RAM 10, while instructing the DSP 9 to reset the internal registers to zero.

After the initialization, the CPU 4 proceeds to the next step SA2, and transfers the speech synthesis program stored in the ROM 5 to the DSP 9. Thus, the DSP 9 is ready for speech synthesis. Next, in step SA3, the value of the counter register i is reset to zero. The value of the counter register i is determined by the speech synthesis parameters (residual Z and Percoll coefficients K1 to Kn).
Is read from the data ROM 8 and set in the RAM 10 or when the speech synthesis parameters stored in the RAM 10 are set in the DSP 9.

In step SA4, the speech synthesis parameters corresponding to the contents of the MIDI data input via the MIDI interface circuit 7 are read from the data ROM 8 and set in the RAM 10 or the input M
Tone information corresponding to IDI data (note-on NON,
A MIDI process (described later) that generates a note-off NOF, velocity data VEL, and pitch frequency PF) and supplies it to the DSP 9 is performed. Next, in step SA5, the panel switch group 2 is scanned to detect a switch operation, and processing corresponding to the detected switch operation is performed. Since then
Steps SA4 to SA4 until the power switch is turned off.
Repeat A5.

(2) Operation of the MIDI Processing Routine When the MIDI processing routine is executed through step SA4 of the main routine, the CPU 4 advances the processing to step SB1 shown in FIG. to decide. Here, if there is no MIDI input, the determination result is “NO”, and this routine is terminated, and the process returns to the main routine. On the other hand, if there is a MIDI input, the determination result is “YES”, and the next step S
The process proceeds to B2. In step SB2, it is determined whether or not the status of the input MIDI data is "note on".

By the way, in the case of this embodiment, the input M
The status of IDI data is "Exclusive",
Either "Note On" or "Note Off". “Exclusive” designates the lyrics associated with each sound in the music by the sound number m in the above-described voice synthesis parameter. In other words, the syllable of the singing voice is first designated by the note number m by "exclusive", then the pitch and volume of the singing voice are designated by "note on", and the sound is muted by "note off".

Therefore, if there is a MIDI input for the first time, the MIDI data becomes "exclusive", and the result of the determination at step SB2 is "NO", and the routine proceeds to step SB3. In step SB3, MI
It is determined whether or not the DI data is “exclusive”. In this case, the determination result is “YES”, and the process proceeds to Step SB4. When proceeding to step SB4, the CPU 4
The tone number m associated with the lyrics (one word) is extracted from the exclusive message in the MIDI data, set in the register exc, and the processing of this routine is completed once.

In this way, note number m is stored in register exc.
Is set, the CPU 4 reads the speech synthesis parameters (residual Z and Percoll coefficients K1 to Kn) corresponding to the tone number m from the data ROM 8 by the processing of a timer interrupt routine described later,
Set to.

Next, when there is a MIDI input of "note on", the result of the determination in step SB2 is "YES".
, And the process proceeds to Step SB5. Step SB
In step 5, the note number included in the MIDI data of "note on" is converted into a pitch frequency PF by the above-described conversion table TBL (see FIG. 2) and set in the DSP 9. Next, when the process proceeds to step SB6, the CPU 4
The velocity value assigned after the note number is extracted from the IDI data, and is extracted from the velocity data V.
EL is set to DSP9. This allows the DSP
At 9, the pitch frequency PF is set in the pulse generator 20, and the velocity data VEL is set in the envelope generator 22.

Then, in step SB7, the counter register i is reset to zero, and in the following step SB8, the envelope generator 22 is instructed to form an attack. As a result, in the DSP 9, the tone synthesis parameters set in the RAM 10, that is, the header DB
Percall synthesis is performed based on the syllable residual Z specified by [exc, i] and the Percoll coefficients K1 to Kn,
A human voice with a pitch based on the pitch frequency PF corresponding to the note number is generated at the volume of the velocity data VEL.

As described above, after the human voice for one word of the lyrics is pronounced as a singing voice, the MID of "note off"
An I input is made. When the MIDI data of "note off" is supplied, the process proceeds to step SB9 via the above-described step SB3, and when the MIDI data of "note off" is supplied, the process proceeds to step SB9.
The result of determination in B9 becomes "YES", and step SB10
Processing proceeds to In step SB10, the CPU instructs the envelope generator 22 to release the sound, and silences the lyrics being uttered.

As described above, in the MIDI processing routine,
When the MIDI data of "exclusive" is input, a note number m associated with the lyrics (one word) is extracted from the exclusive message in the MIDI data and set in the register exc. Hand over note number m. When MIDI data of "note on" is input, the pitch frequency PF corresponding to the note number is set in the pulse generator 20 of the DSP 9 and the velocity data VEL
Is set in the envelope generator 22 of the DSP 9 and the lyrics are pronounced as a singing voice. Subsequently, when MIDI data of "Note Off" is input, the singing voice is released.

(3) Operation of Timer Interrupt Routine The CPU 4 releases the interrupt mask, for example, every 20 msec, executes the timer interrupt routine shown in FIG. 7, advances the process to step SC1, and proceeds with the header DB [ex
c, i] is smaller than “0”, that is,
It is determined whether or not the speech synthesis parameters of the syllable specified by the phone number m have been completed.

Here, if the residual Z is not "-1", the determination result is "NO" and the process proceeds to Step SC2.
In step SC2, the speech synthesis parameters (residual Z and Percoll coefficients K1 to Kn) corresponding to the header DB [exc, i] are read from the data ROM 8 and read from the RAM 10
Set to. As a result, on the DSP 9 side, the RAM 1
Residual Z and Percoll coefficients K1-K set to 0
Percoll synthesis based on n.

Next, at step SC3, the counter register i is incremented to advance, and this interrupt processing is temporarily terminated. Then, at the next interrupt timing, step SC1 is executed. At this time, when the residual Z read out corresponding to the header DB [exc, i] is “−1”, that is, when the end of data is reached, the above-described step SC1 is executed. The result of the determination is "YES", and in this case, the processing is terminated without performing any processing. Therefore, in such a case, since the voice synthesis parameters are not updated on the DSP 9 side, the lattice type filters 25-1 to 25-25 for performing percall synthesis are used.
The time change of −n (see FIG. 4) stops, and the synthesized singing voice enters the sustain state.

As described above, in the present embodiment, the tone information including note-on / note-off, pitch frequency and velocity is set in the DSP 9 based on the MIDI data supplied from the outside, and the tone to be note-on is also set. The voice synthesis parameters of the syllable corresponding to the associated lyrics (one word) are read from the database in the data ROM 8 and set in the RAM 10. As a result, the DSP
In step 9, the human voice which is percall-synthesized according to the given voice synthesis parameter is uttered as a singing voice according to the musical tone information.

In particular, in this embodiment, since the residual Z is set to "-1" in the speech synthesizing parameters of each syllable and an identifier indicating the end of data is provided, the percall coefficients K1 to Kn are set when the end of data is reached. Since the synthesized singing voice is maintained without change with time, the synthesized singing voice becomes a sustained state, and becomes a singing voice with an extended sound (for example, “Ah”). Therefore, a musical tone can be formed as a more natural singing voice.

In this embodiment, the velocity data V
Multiplier 2 so that the amplitude of the excitation waveform can be controlled by EL
3 (see FIG. 4), the lattice type filter 25-1 is provided.
In addition to the formants obtained in 音 25-n, it is also possible to variably control the sound quality (tone intensity).

Further, in the above-described embodiment, the lyrics (one word) associated with the musical sound to be turned on are defined by MIDI.
Since it is defined by an exclusive message of data, it is possible to generate a singing voice in real time.

In addition, in this embodiment, since the envelope is superimposed by the multiplier 23, the volume can be controlled by note-on / off in addition to the coefficient multiplication of the residual Z. Therefore, it is possible to easily control the pronunciation / silence of the singing voice corresponding to the lyrics.

In the above-described embodiment, the pulse generator 20 that outputs a pulse train having a period corresponding to the pitch frequency PF is used as an excitation source for percall-synthesizing a voiced sound. The waveform generation mode shown in FIG. 8 may be used. The waveform generation mode shown in the figure includes waveform generators 30 and 31 for generating waveforms a and b having different waveforms at a pitch corresponding to the pitch frequency PF, and a multiplier 3.
2 and 33 and an adder 34, and an interpolator 35 for interpolating the waveforms a and b according to the velocity data VEL.

According to such a configuration, the waveform a and the waveform b are interpolated according to the velocity data VEL.
It becomes possible to change the timbre according to the velocity value. As the types of the waveforms a and b, those containing many overtone components, such as a triangular wave and waveforms having different pulse widths, are effective. In addition to the above, for example, a plurality of detuned note-ons are used. By simultaneously generating and mixing these waveform signals, voice synthesis such as a chorus can be performed.

[0051]

According to the first aspect of the present invention, the parameter storage means stores the speech synthesis parameters for each of a plurality of syllables, and the music information generating means stores the musical tone information representing each musical tone of the musical composition and each musical tone. When the utterance information indicating the lyrics associated with the lyric is generated, the parameter generating means reads out the speech synthesis parameters of the syllables matching the lyrics from the parameter storage means in accordance with the utterance information. The voice synthesis means synthesizes a human voice based on the voice synthesis parameter output from the parameter generation means and pronounces the voice as a singing voice in accordance with the musical tone information, so that the voice synthesized human voice is formed as a natural singing voice. can do. According to the second aspect of the present invention, the parameter storage means includes:
The speech synthesis parameter for each syllable is provided with an identifier indicating the end of the data. When the speech synthesis means detects this identifier and determines that the speech synthesis parameter has reached the end of the data, the singing voice being synthesized is continuously pronounced. Therefore, a musical tone can be formed as a more natural singing voice. According to the third aspect of the present invention, the voice synthesizer variably controls the sound quality of the synthesized human voice according to the velocity included in the musical tone information, so that it is possible to express the tone of the tone. According to the fourth aspect of the present invention, since the parameter generating means expresses the utterance information by an exclusive message of MIDI data, it is possible to extremely easily generate a singing voice in real time. According to the fifth aspect of the present invention, the voice synthesizing unit controls the singing / sounding of the singing voice according to the note-on / note-off included in the musical sound information. Can be easily controlled.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a tone generator according to the present invention.

FIG. 2 is a diagram for explaining the contents of a conversion table TBL for converting a note number into a pitch frequency PF.

FIG. 3 is a diagram for explaining a data structure of speech synthesis parameters registered in a database in a data ROM 8;

FIG. 4 is a functional block diagram showing a Parcall synthesis algorithm of the DSP 9;

FIG. 5 is a flowchart showing an operation of a main routine.

FIG. 6 is a flowchart showing the operation of a MIDI processing routine.

FIG. 7 is a flowchart showing the operation of a timer interrupt routine.

FIG. 8 is a block diagram showing a modification of the pulse generator 20.

[Description of Signs] 1 keyboard unit 2 panel switch group 3 display unit 4 CPU (parameter generation unit, voice synthesis unit) 5 ROM 6 RAM 7 MIDI interface circuit (song information generation unit) 8 data ROM (parameter storage unit) 9 DSP (Speech synthesis means) 10 RAM (speech synthesis means) 11 D / A converter 12 Amplifier

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁶ Identification code Agency reference number FI Technical display location G10L 5/04 G10H 7/00 513Z

Claims (5)

[Claims] 1. Parameter storage means for storing speech synthesis parameters for each of a plurality of syllables, music information generation means for generating musical tone information representing each musical tone of a musical composition and utterance information representing lyrics associated with each musical tone. A parameter generating means for reading out and outputting a voice synthesizing parameter of a syllable matching the lyrics from the parameter storage means in accordance with the utterance information generated by the music information generating means; A musical sound generating device comprising: a voice synthesizing unit that generates a human voice as a singing voice according to the musical sound information. 2. The parameter storage means includes an identifier indicating the end of data in a speech synthesis parameter for each syllable. The speech synthesis means detects the identifier and determines that the speech synthesis parameter has reached the end of data. 2. The musical sound generator according to claim 1, wherein when it is determined, the singing voice being synthesized is continuously generated. 3. The musical sound generating apparatus according to claim 1, wherein said voice synthesizing means variably controls the sound quality of the synthesized human voice according to the velocity included in the musical sound information. 4. The musical sound uttering device according to claim 1, wherein said parameter generating means expresses said utterance information by an exclusive message of MIDI data. 5. The musical sound producing apparatus according to claim 1, wherein said voice synthesizing means controls the pronunciation and mute of a singing voice by note-on note-off included in said musical sound information.