JPH0364880B2 - - Google Patents

Description

[Detailed Description of the Invention] The present invention relates to an electronic musical instrument that generates human voice sounds;
In particular, the present invention relates to an electronic musical instrument in which the phonology of human voice sounds to be generated is changed depending on the length of key press time.

Recently, attempts have been made to generate a human voice, that is, human voice sound, in an electronic musical instrument and use it as the backing chorus of a song. This type of electronic musical instrument forms a fixed formant filter using various methods and uses it to generate human voice sounds, and the phonology of the generated human voice sound is set by appropriately selecting the above filters. be done. By the way, listening to human voices with the same phoneme repeatedly feels monotonous, and the phoneme tends to become difficult to understand as you listen.

Therefore, it is possible to switch the tone in the middle of a song, but this would require operating a tone selection switch each time, which would be very cumbersome and undesirable when playing an electronic musical instrument. .

The present invention has been developed in consideration of the above points, and is designed to detect the length of key press time and change the phonology of human voice sounds in accordance with the key press length.

The present invention will be explained below based on the drawings.

FIG. 1 is a basic configuration block diagram showing an embodiment of an electronic musical instrument according to the present invention. The keyboard section 1 has, for example, an upper keyboard UK, a lower keyboard LK, and a pedal keyboard PK, and key press operations on each of the keyboards UK, LK, and PK are detected by a key press detection section 2. The pressed key detection section 2 detects a key being pressed on the keyboard section 1, and outputs key information (hereinafter referred to as a key code KC) representing the pressed key. The sound generation assignment section 3 assigns each pressed key on the keyboard section 1 to one of a plurality of musical tone generation channels based on the key code KC added from the pressed key detection section 2, and assigns each pressed key to one of a plurality of musical tone generation channels. A key code KC indicating a key and a key-on signal KON indicating whether the key is kept pressed or released are output in a time-division manner in synchronization with the channel timing of each channel. Here, the key code KC is, for example, the keyboard to which the pressed key belongs (UK, LK, or
PK), 2-bit keyboard codes K ₂ , K ₁ , 3-bit octave codes B ₃ , B ₂ , B ₁ indicating the octave range of the pressed key, and 4-bit note code N ₄ , indicating the note name. This is 9-bit data consisting of N ₃ , N ₂ , and N ₁ . In addition, the key-on signal KON
is a signal that is "1" while the key press continues, and becomes "0" when the key press is released (the key is released). The key code KC and key-on signal KON outputted from the sound generation assignment section 3 are supplied to the general musical tone signal forming section 5 and the vocal sound detecting section 6.

The general musical tone signal forming section 5 forms musical tone signals corresponding to the keys assigned to each channel in response to the key code KC and the key-on signal KON given in a time-sharing manner from the sound generation assignment section 3,
The musical tone signals of each channel thus formed are mixed appropriately and then outputted to the mixing section 8. Further, the vocal sound detection section 6 detects a key that generates a human voice sound (vocal sound) from among the keys pressed on the keyboard section 1, and the vocal sound detection section 6 detects a key that generates a human voice sound (vocal sound) from among the keys pressed on the keyboard section 1. Based on KC and key-on signal KON, one or more predetermined keys are selected from among the keys currently being pressed, and vocal sound key code VKC and vocal sound key-on VKON are output corresponding to the selected keys to form a vocal musical sound signal. 7. This vocal sound detection section 6 is configured as shown in FIG. 2, for example.

In the embodiment of the vocal sound detection section 6 shown in FIG. 2, a predetermined keyboard (UK, LK, or PK)
is configured to detect a single key corresponding to the highest pressed key among the keys currently being pressed, and output a vocal sound key code VKC and a vocal sound key-on signal VKON in response to this highest pressed key. ing.

In FIG. 2, among the key codes KC and key-on signals KON of each channel given in a time-sharing manner from the sound generation allocation section 3, the keyboard codes K ₂ and K ₁ are input to the decoder 61, and the octave codes B ₃ to _{B 1} are input to the decoder 61. , note codes N ₄ to N ₁ and key-on signal KON are input to the gate 62 . The decoder 61 decodes the keyboard codes K ₂ and K ₁ and outputs an upper keyboard signal U, a lower keyboard signal L, and a pedal keyboard signal P indicating the keyboard to which the keys assigned to each channel belong, and these signals U, L and P are supplied to AND gates A ₁ , A ₂ and A ₃ respectively. The output signals of the upper keyboard selection switch S _U , the lower keyboard selection switch S _L , and the pedal keyboard selection switch _SP are applied to the other inputs of the AND gates A ₁ , A ₂ , and A ₃ . In this case, the selection switches S _U , S _L , and _SP are used to select which of the pressed keys of the keyboard is to be detected when the vocal sound detecting section 6 detects the highest pressed key. For example, when the selection switch S _U is turned on, the output signal of the switch S _U becomes "1", which enables the AND gate A ₁ to operate and converts the upper keyboard signal U output from the decoder 61 to the OR gate G ₁ is added to the enable terminal E of the gate 62 via the gate 62 to make the gate 62 conductive. Therefore, in this case, the key code KC of the channel to which the key of the upper keyboard UK is assigned is
When the key-on signal KON is applied, the upper keyboard signal U (“1” signal) is outputted from the decoder 61 at the channel timing of this channel, thereby making the gate 62 conductive and outputting the octave codes B ₃ to _{B 1} of the channel. , note code
Pass N ₄ to N ₁ and key-on signal KON.
That is, only the octave codes B ₃ to _{B 1} , the note codes N ₄ to _{N 1} and the key-on signal KON relating to the upper keyboard UK channel are selected by the gate 62 and applied to the highest note detection circuit 63 . Also, when the selection switch S _L or _SP is turned on, the octave codes B ₃ to B 1 and note code N ₄ of the channel to which the keys of the lower keyboard LK or pedal keyboard PK are assigned are set in the same way as _above . ~ _N1 and the key-on signal KON are selected by the gate 62 and applied to the highest note detection circuit 63. In addition, the selection switch
When multiple switches are turned on at the same time in S _U , S _L , and _SP , octave codes B ₃ to B ₁ , note codes N ₄ to _{N 1} , and key on keys for each keyboard channel corresponding to the selected switch turned on are displayed. signal
All KONs are selected and added to the highest pitch detection circuit 63.

The highest pitch detection circuit 63 detects the octave codes B ₃ to _{B 1} and note codes added from the gate 62.
N ₄ to N ₁ and key-on signal KON (selection switch
The octave code corresponding to the highest note of the currently pressed key among S _U , S _L , and S _P (related to the keyboard corresponding to the switch that is turned on)
B ₃ to _{B 1} and note codes N ₄ to _{N 1} are detected. This detection is performed by sequentially comparing the added octave codes B ₃ to _{B 1} and note codes N ₄ to _{N 1} while the channel timing of each channel goes through one cycle, and the highest note pressed key detected as a result of the comparison is Octave code B ₃ ~ B ₁ , note code N ₄ ~
_N1 is stored for one cycle of the next channel timing and output as the vocal sound key code VKC. In addition, in the highest note detection circuit 63, a vocal sound key-on signal VKON is formed in connection with the above-mentioned highest note pressed key detection, and this signal
VKON is a vocal sound key code VKC that has some value other than "0" when the highest pressed key is detected.
When VKC is being output, it becomes "1" indicating a pressed key, and when the highest pressed key is no longer detected and the content of the key code VKC becomes "0", it becomes "0" indicating a key release. Note that this vocal sound key-on signal VKON becomes "0" for a predetermined short time when the detected highest-pitched pressed key is changed (that is, when the content of the key code VKC is changed).
This is achieved by setting the key-on signal VKON to "0" for a short period of time to switch the vocal tone signal formed in the vocal tone signal forming section 7, which will be described later, from one related to the old highest pressed key to a new one related to the highest pressed key. This is to make the division clear.

In this way, the vocal sound key code VKC and the vocal sound key-on signal VKON output from the vocal sound detection section 6 are supplied to the vocal musical sound signal formation section 7.

The note length detection unit 4 detects the length of the highest note pressed key detected by the vocal sound detection unit 6, that is, the note length of the highest note pressed key, and outputs a note length signal B. is shown in Figure 3.

In this embodiment, the note length of the highest pressed key is detected by measuring the time during which the vocal sound key-on signal VKON output from the vocal sound detection section 6 is in the "1" state. That is, the vocal sound key-on signal VKON is delayed for a short time by the delay circuit D and applied to one input of the AND gate _A4 , and its rising edge is differentiated by the differentiating circuit 41 and applied to the reset input R of the counter 42. In addition, vocal sound key-on signal
The rising edge of VKON is differentiated by a differentiating circuit 43 and applied to the latch control input L of the latch circuit 44. Now, when the vocal sound key-on signal VKON is supplied as _shown in FIG. Therefore, the count value of the counter 42 increases sequentially while the vocal sound key-on signal VKON is "1" as shown in FIG. Change stops. And then the vocal sound key-on signal
When VKON becomes "1", the count value of the counter 42 is changed to the latch circuit 4 by the output of the differentiating circuit 43.
4 (see FIG. 4C), and immediately after that, the counter 42 is reset. In this way, the latch circuit 44 latches data corresponding to the length of the key pressed at the highest note, that is, the note length, and outputs it as a note length signal B.

The vocal tone signal forming section 7 generates a vocal tone key code VKC and a vocal tone key-on signal.
It forms and outputs a musical tone signal corresponding to a predetermined human voice sound based on VKON, and the phoneme of the formed musical tone signal is a note length signal B given from the note length detection section 4.
Changed by The vocal musical tone signal forming section 7 generates musical tone signals corresponding to various phonemes of human vocal sounds in a predetermined order or in an arbitrary order based on the note length signal B output from the note length detecting section 4,
If the length of the highest pressed key (note length signal B) on the keyboard section 1 changes, the phoneme will change accordingly. Note that this musical tone signal forming section 7
The pitch (pitch) of the musical tone signal formed by the above corresponds to the highest pressed key indicated by the key code VKC.

The musical tone signal of the human voice outputted from the vocal musical tone signal forming section 7 is supplied to the mixing section 8, where it is mixed with the musical tone signal corresponding to each pressed key outputted from the general musical tone signal forming section 5, and then an amplifier is used. , are input to a sound system 9 consisting of speakers and the like, and are produced as musical tones.

Next, the specific configuration and operation of the vocal tone signal forming section 7 will be explained.

FIG. 5 shows an example in which the vocal tone signal forming section 7 is configured using a digital filter. The decoder 71 decodes the key press length of the highest note being pressed at the time from the note length signal B output from the note length detector 4, and selects a predetermined phoneme corresponding to the decoded value. The selection signal S is sent to the address signal generator 7.
2 and selector 74. In this case, the note length signal B output from the note length detection section 4 is a digital value, and the value changes frequently. Therefore, it is preferable that the decoder 71 decodes only the upper predetermined bits of the digital value as a note length signal (that is, ignores the lower bits) and outputs the phoneme selection signal S corresponding to the decoded value.

Instead of the decoder 71, a phoneme switching control circuit as shown in FIGS. 6 to 8 may be used.

The phoneme switching control circuit 171 in FIG. 6 includes a comparison circuit 171a, an M-ary counter 171b, and a decoder 171c, and compares the note length signal B from the note length detection section 4 with a predetermined value K by the comparison circuit 171a. Then, when B≦K (that is, the detected symbol length is less than or equal to a predetermined length), a phoneme switching signal P is output, this phoneme switching signal P is counted by the M-ary counter 171b, and the decoder 171c outputs the phoneme switching signal P to the count value of the counter 171b. Corresponding phoneme selection signals S are output in a predetermined order (for example, in the order of A, U, O, La, and Ru). When the phoneme selection signal S changes by a predetermined number of phonemes, it returns to the initial phoneme selection signal S again and the same pattern is repeated thereafter.

The phoneme switching control circuit 172 in FIG. 7 includes a comparison circuit 172a having the same configuration as the comparison circuit 171a of the embodiment shown in FIG. 6, and a random number generator that generates numbers "1" to "M" in an arbitrary order. 172b and comparison circuit 1
It consists of a latch circuit 172c that latches random number data generated from a random number generator 172b when it receives a phoneme switching signal P from 72a, and a decoder 172d that decodes the random number data latched and output by the latch circuit 172c. When the length of becomes equal to or less than a predetermined length, the random number generator 172b outputs a phoneme selection signal S corresponding to the randomly generated random number data. In this case, the order in which the phonemes occur is completely random.

In the phoneme switching control circuit 173 in FIG. 8, the comparator circuit 173a outputs "1" when the note length signal B output from the note length detector 4 is smaller than the predetermined length K, and "0" when it is larger. Phoneme switching signal P
is output. When the phoneme switching signal P is "1", the selector 173b selects and outputs the signal added to its A input, that is, the signal obtained by differentiating the rising edge of the vocal sound key-on signal VKON with the differentiating circuit 173c, and outputs the phoneme switching signal. When P is "0", the signal applied to the B input, that is, the signal obtained by frequency-dividing the output of the differentiating circuit 173c by the frequency dividing circuit 173d, is selected and output. counter 1
73e counts the output of selector 173b,
The count value is decoded by a decoder 173f and output as a phoneme selection signal S corresponding to the count value. As a result, when the key press length, that is, the note length is shorter than the predetermined length, the count value of the counter 173e increases at a faster rate, so that the phoneme is switched more quickly than in the opposite case.

Returning to FIG. 5 again, the address signal generator 7
2 is this phoneme selection signal S and the vocal sound detection section 6
Vocal sound key-on signal output from VKON
Based on this, an address signal ADR for reading the coefficient data stored in the coefficient memory 73 is output for each phoneme. An embodiment of this address signal generating section 72 is shown in FIG.
This will be explained based on FIG. 0 and FIG.

In FIG. 9, when the phoneme selection signal S is input to the address signal generation section 72, the attack pulse generation circuit 10 generates an attack pulse AP for regulating the rise state of human voice sounds, and the decay pulse generation circuit 11 generates an attack pulse AP for regulating the rise state of human voice sounds. A decay pulse DP regulating the state is generated, and a sustain address A _S is generated from the sustain address memory 12 to specify an address for reading coefficient data from the coefficient memory 73 for forming a sustained part of a human voice sound. In this case, attack pulse AP
The period of the decay pulse DP is set according to the phoneme specified by the phoneme selection signal S, and the attack pulse generation circuit 10 and the decay pulse generation circuit 11, for example,
It consists of an oscillator whose oscillation frequency is controlled by Further, the sustain address memory 12 stores the above-mentioned sustain address A _S at each address corresponding to each phoneme, and when the phoneme selection signal S is input as an address signal, the phoneme specified by the signal S is selected. Supported sustain address
Read _AS .

Now, when the vocal sound key-on signal VKON as shown in FIG. 10A is input from the vocal sound detection section 6 to the address signal generation section 72,
The rising edge is differentiated by the key-on pulse.
KNP is output. The counter 17 is reset by this key-on pulse KNP, and RS
Flip-flop 15 is set. As a result, the AND condition of AND gate A ₄ is satisfied and the attack pulse AP is connected to AND gate A ₄ and OR gate.
Since the attack pulse AP is input to the counter 17 via _G4 , the counter 17 starts counting the attack pulse AP. At this time, since the RS flip-flop 16 of the AND gate _A5 has been reset as will be described later, the AND condition is not satisfied and the decay pulse DP is not supplied to the counter 17. FIG. 11 shows the change in the count value of the counter 17 over time, and the counter 17 receives the attack pulse.
AP is counted, and the count value is sent to the coefficient memory 73 as an address signal ADR.
When the count value of the counter 17 becomes equal to the sustain address _A3 output from the sustain address memory 12, a match signal EQ is output from the comparator 18, and the RS flip-flop 15 is reset. As a result, the AND condition of AND gate _A4 no longer holds, and the attack pulse
AP is not supplied to counter 17 and counter 17
The count value remains stopped at the sustain address A _S. This count value (A _S ) continues to be output as the address signal ADR. This state continues until the vocal sound key-on signal VKON falls. When the vocal sound key-on signal VKON falls, the differentiation circuit 14 detects the fall and outputs a key-off pulse KFP. As a result, the RS flip-flop 16 is set and the AND gate is
The AND condition of _A5 is satisfied, and the decay pulse DP generated from the decay pulse generation circuit 11 is input to the counter 17. As a result, counter 17
is the decay pulse from the above count value (A _S )
Performs DP counting operation. During this time, counter 1
The count value of 7 is also sent to the coefficient memory 73 as the address signal ADR. Thereafter, when the count value of the counter 17 reaches the maximum amount N (all bits are "1"), the AND condition of the AND gate _A6 is satisfied and the RS flip-flop 16 is reset. As a result, the input of the decay pulse DP to the counter 17 is stopped, the count value of the counter 17 remains at the maximum amount N, and this count value N is used as the address signal ADR to the coefficient memory 7.
3 continues to be sent.

Returning to FIG. 5 again, human voice sounds are generally created by synthesizing voiced and unvoiced sounds and passing them through a fixed formant filter. Therefore, the coefficient memory 73 stores phonemes that are desired to be generated as human voice sounds. (For example, "a", "u", ... "ru", "wa", etc.)
A memory 73 stores amplitude coefficients of voiced and unvoiced sounds necessary to form musical tones, and Parcor coefficients for determining formalman characteristics for each phoneme.
a, 73b, . . . in advance. Therefore, when the address signal ADR as shown in FIG.
The voiced amplitude coefficients, unvoiced amplitude coefficients, and percoll coefficients shown in FIG. read out,
The signals are supplied to input terminals 1 to M of the selector 74, respectively. The selector 74 selects a coefficient (unvoiced amplitude coefficient, unvoiced amplitude coefficient, Select and output only the voiced amplitude coefficient and Percoll coefficient. In this case, the unvoiced amplitude coefficient is always set to zero for voiced sounds, such as the phoneme of "a," and for cases where an unvoiced sound is mixed at the beginning of the sound, such as the phoneme of "pa" or "ta," the unvoiced amplitude coefficient is always set to zero. If so, the voiced amplitude coefficient increases slowly over time, as shown in Figure 10C, and the unvoiced amplitude coefficient has a sharp peak only for a relatively short period of time at the first rise, as shown in Figure 10D. Shisonogo is set to zero. Note that the Percoll coefficient is set so as to be convenient in terms of time at the rising and falling parts of the sound, as shown in FIG.

Of the voiced sounds and unvoiced sounds forming the human voice, the unvoiced sound signal is generated by multiplying the noise signal output from the noise sound source 75 serving as the sound source by the unvoiced sound amplitude coefficient output from the selector 74 in a multiplier 76.

On the other hand, a voiced sound signal is generated as follows. First, the frequency number F (constant) corresponding to each key of keyboard section 1
The frequency number F stored at the address specified by the vocal sound key code VKC is read out from the frequency information memory 77, which is constituted by a ROM, for example, and which has been stored in advance at each address. The frequency number F read from the frequency information memory 77 is sequentially accumulated at fixed time intervals in the accumulator 78, and the accumulated value is stored in the sound source waveform memory 7.
It is output as a 9 address signal. In the sound source waveform memory 79, the amplitude value at each sample point obtained by dividing a basic waveform forming a voiced sound (for example, one period of an asymmetric triangular wave) into a plurality of sample points is stored as digital data at each address. The amplitude values are sequentially read out based on the address signal output from the accumulator 78.
The voiced sound source signal read out from the sound source waveform memory 79 is multiplied by the voiced amplitude coefficient output from the selector 74 in a multiplier 80 to become a voiced sound signal. The thus formed voiced sound signal is sent to a multiplier 76.
The adder 81 adds the signal to the unvoiced sound signal output from the adder 81 and supplies the resultant signal to the digital filter 82 . The digital filter 82 is, for example, a
As disclosed in No. 56-125798, a phonetic formant is formed by a plurality of filters each composed of a plurality of multipliers, an adder, and a memory, and the filter characteristics are determined by the percoll from the selector 74. A human voice sound signal (vocal musical sound signal) corresponding to a predetermined phoneme is formed under control using coefficients. The digital human voice signal output from the digital filter 82 is converted into an analog vocal tone signal by the D/A converter 83.

In this way, the vocal tone signal generated from the vocal tone signal forming section 7 is mixed with the musical tone signal generated from the general musical tone signal forming section 5 in the mixing section 8, as shown in FIG. pronounced.

In addition to the above-mentioned digital filter method, the vocal tone signal forming section 7 also uses (1) a formant filter by combining a plurality of analog bandpass filters, and adjusts the center frequency and resonance characteristics of each bandpass filter to each phoneme; Analog filter method (for example, Japanese Patent Application Laid-Open No. 55-77799)
(2) A harmonic synthesis method that uses a harmonic component generation circuit to control the amplitude of each harmonic component in accordance with a predetermined phoneme (for example,
(Refer to Publication No. 21063), (3) A plurality of formant forming circuits that perform frequency modulation calculations are provided, and the center frequency, level, and spread of the formant of each formant forming circuit are controlled for each phoneme to control the formant characteristics of one phoneme. A frequency modulation method (for example, Japanese Unexamined Patent Application Publication No. 18623/1983) can be used. In addition, as a means for generating a musical tone signal of a human voice, the waveform of the human voice of a desired phoneme itself may be directly stored in a waveform memory and read out. In this case, methods for storing the human voice waveform in the waveform memory include PCM (pulse code modulation), APCM (adaptive pulse code modulation),
DPCM (Differential Pulse Code Modulation), ADPCM (Adaptive Differential Pulse Code Modulation), DM (Delta Modulation), LPCM
There are various methods such as (linear pulse code modulation), and any of them may be used. Furthermore, in such a storage system, a more natural human voice can be obtained by providing a waveform memory that stores the above-mentioned human voice sound waveform for each pitch.

In the above-mentioned embodiment of the electronic musical instrument according to the present invention, only one sound corresponding to the highest pressed key of each pressed key is used as the human voice sound, but instead of the highest pressed key, the human voice sound corresponding to the lowest pressed key is used. Alternatively, human voice sounds may be generated for each key of the upper keyboard, lower keyboard, and pedal keyboard, or for all pressed keys. Furthermore,
The range of the human voice to be produced may be different from the range of the pressed key (for example, the human voice may be produced in a range one octave above or below the range of the pressed key).
Furthermore, in the above embodiment, the general musical tone signal forming section may be omitted as appropriate to generate only human voice sounds. Furthermore, the key press length (note length) may be detected based on the output of the key press detection section 2 or the output of the pronunciation assignment section 3. Furthermore, instead of the sound generation assignment method, an indirect opening/closing method can be used to process the key depression information from the keyboard section, immediately after directly processing signals caused by the on/off of switches provided for each key.

In addition, the order of the phonemes of human voice sounds to be pronounced is programmed in advance and stored in a memory, etc., and the contents of the memory are sequentially read out using a phoneme switching signal that is output based on the length of the key pressed, that is, the note length. The phoneme of the human voice may be switched.
In this case, the phonological program of the human voice may be set arbitrarily by the player of the electronic musical instrument.
Note that the present invention can also be realized in an electronic musical instrument whose keyboard section consists of a single-stage keyboard.

As described above, in this invention, the phoneme of the human voice sound to be pronounced is switched according to the length of the key pressed, that is, the note length of the note, so the phoneme of the human voice sound can be automatically changed without the need for troublesome switch operations. This makes it possible to produce human voice sounds that are rich in variety, without the monotony of conventional methods.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing a schematic configuration of an electronic musical instrument according to the present invention, FIG. 2 is a block diagram showing a circuit configuration of an embodiment of the vocal sound detection section shown in FIG. 1, and FIG. 1 is a block diagram showing the circuit configuration of an embodiment of the note length detection section shown in FIG. 1, FIG. 4 is a time chart explaining the operation of the note length detection section shown in FIG. 3, and FIG. Block diagrams showing one embodiment of the vocal tone signal forming section shown in FIGS. 6, 7, and 8 are used in place of the decoder used in the vocal tone signal forming section shown in FIG. 5. FIG. 9 is a block diagram showing an example of an address signal generating section used in the vocal tone signal forming section shown in FIG. 5; 1st
0 and 11 are explanatory diagrams of the operation of the address signal generating section shown in FIG. 9. 1...Keyboard section, 2...Key press detection section, 3...Pronunciation assignment section, 4...Note length detection section, 5...General musical tone signal forming section, 6...Vocal sound detection section, 7...Vocal Musical tone signal forming section, 8... Mixing section, 9... Sound system, 71... Decoder, 171, 17
2,173... Phoneme switching control circuit, 72... Address signal generation section, 73... Coefficient memory, 74...
Selector, 83...Digital filter.

Claims (1)

[Scope of Claims] 1. A keyboard section, a pressed key detection means for detecting a key pressed on the keyboard section and outputting key information representing the key, and detecting a key pressed on the keyboard section based on the key information. a key pressing length detecting means for detecting the length of time; a human vocal sound forming means for forming a musical sound signal of a human vocal sound based on the key information; 1. An electronic musical instrument characterized by comprising: a phoneme switching means that switches according to the length of the key press time detected by the key press length detecting means. 2. The electronic musical instrument according to claim 1, wherein the phoneme to be switched by the phoneme switching means is predetermined corresponding to each length of key press time. 3. The human voice musical tone forming means selects one or more predetermined pieces of key information from the key information and forms a musical tone signal corresponding to the selected key information. The electronic musical instrument according to item 2.