JP3235315B2 - Formant sound source - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a sound source for electrically synthesizing formant sounds, such as human voice sounds and natural musical instrument sounds, and more particularly to a formant sound source capable of synthesizing formant sounds close to natural sounds. It is.

[0002]

2. Description of the Related Art A general method of generating a formant waveform will be described with reference to FIG. Thereby generating a periodic waveform having a certain frequency f _f as shown in FIG. 14 (a), generating a window function signal that a smooth change as shown in FIG. (B). By the window function signal, the amplitude modulates a periodic waveform having the frequency f _f, the waveform becomes a formant waveform as shown in FIG. (C). As the formant waveforms is shown, the envelope is the same shape as the window function, the carrier signal is a periodic waveform having a frequency f _f, the frequency f _f is the center frequency of the formant.

Such a formant waveforms, for example, is generated every period t _f as shown in (c), the period t _f is to be a pitch period. Further, the spectrum of the formant waveforms, for example, as shown in FIG. 3 (b), there is a relatively sharp spectrum distribution having a peak at frequency f _f of the periodic signal. In order to obtain various timbres using the formant waveform generated in this manner, conventionally, various shapes of the window function waveform are set, and an example of such a formant type sound source is shown in FIG.
It is shown in FIG. The formant type sound source shown in FIG. 1 is described in Japanese Patent Application Laid-Open No. 2-269898 filed by the applicant of the present application.

In FIG. 15, a pulse generation circuit 101
Is a phase accumulator 111 and a differentiator 112
Consists of The phase accumulator 111 synchronizes with a predetermined clock pulse to generate a formant fundamental frequency information value F.
₀ has been accumulated. This phase accumulator 111
FIG. 16 (a) shows the accumulated value output. The accumulated value changes in a saw-tooth shape, and overflows every time the value reaches 2π, and is again accumulated. by the way,
The fundamental frequency information value F ₀ is set to a value corresponding to the fundamental frequency of the formant sound to be generated, that is, a value corresponding to the pitch, and the phase accumulator 111 sets the accumulated value qF
Each time ₀ (q = 1, 2, 3,...) Reaches a value corresponding to 2π, a signal “1” is output as the most significant bit (MSB). Since the differentiated pulse shown in FIG. 16B is generated by differentiating the MSB signal by the differentiator 112, the period of the differentiated pulse is equal to the period of the pitch to be generated. This differentiated pulse is supplied to the carrier wave generation circuit 102 and the modulation wave generation circuit 103 as an initial setting signal.

[0005] The carrier generation circuit 102 comprises a phase accumulator 121 and a sine wave table 122.
The phase accumulator 121 accumulates the formant center frequency F _f in synchronism with the clock pulses, and sequentially outputs the accumulated value qF _f as a read address signal of the sine wave table 122. In the sine wave table 122,
One cycle of the sine wave amplitude is sequentially stored at each address as a logarithmic value, and the sine wave stored at the address specified by the address signal (accumulated value qF _f ) output from the accumulator 121. The amplitude values are sequentially read. Thus, sequential sample points amplitude value of the sine wave of the formant center frequency corresponding to the frequency signal F _f is the sine wave table 122 are sequentially output according to the clock pulses.

The formant center frequency information value F _f
Is the pitch f _o of to be generated formant sound independently is set to a value corresponding to the formant center frequency f _f is one of the parameters indicating the tone of the formant sound. Accordingly, the carrier wave generating circuit 102 to generate a sine wave of frequency equal to arbitrarily set the formant center frequency f _f in independently corresponding to a desired tone color from the tone pitch of the to be generated formant sound Become. The modulation wave generating circuit 103 is composed of selectors 131 and 134, a phase accumulator 132, a sine square wave table 133, and a bit shifter 135. The selector 131 operates at a value where the accumulated value output of the phase accumulator 132 corresponds to 0 to π. , The bandwidth value ka for the first half cycle is selected and supplied to the phase accumulator 132, and when the accumulated value output of the phase accumulator 132 is a value corresponding to π to 2π, the bandwidth value kb for the second half cycle is selected and the phase accumulator is selected. 132. When the accumulated value output of the phase accumulator 132 is a value corresponding to 0 to π, the selector 134 sets the shift value n for the first half cycle.
a is selected and supplied to the bit shifter 135, and when the accumulated value output of the phase accumulator 132 is a value corresponding to π to 2π, the second half cycle shift value nb is selected and supplied to the bit shifter 135.

The phase accumulator 132 accumulates the bandwidth ka or the bandwidth kb selected by the selector 131 in synchronization with the clock pulse, and accumulates the accumulated value q ₁ ka + q ₂ kb (q ₁ = 1, 2, 3, ..., q
₂ = 1, _2, 3,...) To the sine square wave table 133
Are sequentially output as a read address signal. When the accumulated value reaches a value corresponding to 2π, the phase accumulator 132 stops the accumulating operation and continues to output a value corresponding to 2π, as shown in FIG. Further, the accumulator 132 is set so that the most significant bit (MSB) becomes “1” when the accumulated value is equal to or more than π, as shown in FIG. The signal is supplied to selectors 131 and 134 as a select signal.

[0008] In the sine square wave table 133, the amplitude values of successive sample points of one cycle of the sine square wave are stored in each address as logarithmic values, and an address signal (accumulated value q ₁₎ output from the accumulator 132 is stored. When the stored sample values of the sine square wave are sequentially read from the address specified by (ka + q ₂ kb), the waveform becomes as shown in FIG. The bit shifter 135 shifts the sine square wave sample value read from the sine square wave table 133 by the number of bits corresponding to the shift value na or the shift value nb supplied through the selector 134 to the left, and as a result, The sample value is multiplied by 2 ^na or 2 ^nb . As a result, a waveform of 2 ^{na of the} sine function or 2 ^nb of the sine function is obtained. Therefore, if the shift value na and the shift value nb are set to different values, a sine-powered waveform having a different power between the first half cycle and the second half cycle is obtained, and therefore, as shown in FIG. And the shape of the waveform in the latter half cycle becomes different.

The sine-powered waveform signal generated in this manner is input to one input terminal of adder 104,
To the other input terminal of the adder 104, the the signal of the generated formant center frequency f _f from carrier wave generating circuit 102 is input from the adder 104, the signals obtained by adding the two input signals are output Will be done.
The output of the adder 104 is further input to one of the inputs of the adder 105, and is added to the envelope signal ENV which is logarithmic value data. Then, when the output of the adder 105 is input to the logarithm / antilogarithm converter 106, the added logarithmic data is converted into an antilog value. Then, since the addition of logarithms is equivalent to the multiplication of the true number of the Signs powers waveform signal from the bit shifter 135 is added by the adder 104, the signal of the carrier wave generated from the generating circuit 102 formant center frequency f _f Is multiplied by the envelope signal ENV added by the adder 105 to output a signal obtained by multiplying the multiplied signal by the adder 105.

[0010] That is, the amplitude-modulated signal by sine power of waveform signal a signal of the formant center frequency f _f output from the bit shifter 135 (e.g., FIG. 14
(See (c)), and the amplitude-modulated signal is further subjected to amplitude modulation by the envelope signal ENV.
The amplitude-modulated signal output from the logarithm / antilogarithm converter 106 is converted to a digital / analog (DA) converter 10.
The signal is converted into an analog signal by a signal generator 7 and is emitted as a formant sound through a sound system including an amplifier and a speaker system (not shown).

When controlling the skirt characteristic of the formant waveform in this formant type sound source,
What is necessary is just to change the shape of the modulation wave generated from the modulation wave generation circuit 103. This modulated wave is generally called a window function, and changes the bandwidth values Ka and Kb that determine the time width of the formant supplied to the selector 131 and the shift values na and nb supplied to the selector 134. This changes the window function. For example, the spectral distribution of the formant sound modulated by a window function generated when the bandwidth value K is K = Ka = Kb = 2π * 100 and the shift values are na = 0 and nb = 1 is shown in FIG.
As shown in FIG. Also, without changing the bandwidth value K,
According to the window function when the shift value is changed to na = 0, nb = 4, the spectrum becomes a formant sound as shown in FIG. Further, the bandwidth value K is calculated as K = Ka = Kb = 2.
According to the window function when π * 400 is set and the shift value is changed to na = nb = 0, a spectrum formant sound as shown in FIG. 19 is obtained. Thus, changing the window function by changing the bandwidth value or shift value,
I try to create various formant sounds.

[0012]

However, if the window function is changed to change the skirt characteristic of the formant by changing the band width value or shift value that determines the time width of the formant as described above, FIG. As shown in FIG. 19, there is a problem that the bandwidth of the spectral distribution characteristic also changes at the same time (as the skirt characteristic expands, the bandwidth also changes). Further, since the shift value is an integer, the control of the skirt characteristic by changing the shift value is extremely coarse control, and there is a problem that the hardware is extremely complicated if the control is performed finely. . Accordingly, an object of the present invention is to provide a formant type sound source that can control the skirt characteristic of the spectrum distribution with the formant waveform without changing the bandwidth.

[0013]

In order to achieve the above object, the present invention provides a formant by changing the time of the first half cycle of the window function with respect to the time width of the window function without changing the time width of the window function. And a window function time width generating means for setting the time width of the window function and the first half cycle time width of the window function according to the timbre information, and the window function time width generating means. Window function generating means for generating a window function signal having the set time width and the first half cycle time width characteristic for each pitch cycle in accordance with the output window function information and the pitch information; and And a periodic function generating means for generating a formant center frequency signal, wherein the window function signal generated by the window function generating means generates the formant center frequency signal. By amplitude modulating the formant center frequency signal, it is obtained so as to generate a formant sound.

[0014]

According to the present invention, by changing the first half period time of the window function with respect to the time width, the skirt characteristic of the formant can be changed without changing the bandwidth, and control can be performed. Hardware can be simplified. Further, since the first half cycle time with respect to the time width of the window function can be continuously changed, the skirt characteristic can be finely controlled. In addition, if the window function is a function based on a sine wave, the convergence is fast, so that the possibility that formant waveforms are superimposed can be prevented.

[0015]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Prior to the description of the present invention, the principle of generating a window function in the present invention will be described with reference to FIGS. In the graph shown in FIG. 1, the horizontal axis is time,
The vertical axis is the phase. This phase corresponds to the address of the waveform memory where at least a part of the window function waveform is stored. Also, t _win is the time width of the window function, t
_sk is the first half cycle time of the window function. Therefore, the latter half cycle time of the window function is (t _win -t _sk ).

That is, the time of the first half cycle of the window function is t _sk
When t _sk is set to be shorter as shown in FIG. 1A, the first half period portion of the window function read within this t _sk is read. Are read out after being compressed as shown in FIG. On the other hand, the time of the latter half cycle of the window function is (t
_win− t _sk ), and the latter half of the window function read out within (t _win −t _sk ) is expanded and read out as shown in FIG. Therefore, a window function is read in which the first half period portion rises sharply and the second half period portion gradually falls as shown in FIG. In the waveform memory from which the window function is read, for example, si
The waveform of n ² θ is stored.

The time _tsk of the first half cycle of the window function is １ ／ t
Figure 2 shows the change of the window function when changing from _win to 1 / 32t _win
Shown in In this figure, (1) is t _sk = 1 / 2t _win
Is when you set, (2) is the time when you set _{t sk = 1 / 4t win,} (3) is when the set with _{t sk = 1 / 8t win,} (4) is t _sk = is when you set 1 / 16t _win, (4) is when you set t _sk = 1 / 32t _win. As described above, as the time _tsk of the first half period of the window function becomes shorter, the first half period portion of the window function rises sharply. Next, FIG. 3 to FIG. 8 show a formant waveform generated by modulating a formant center frequency signal composed of a periodic waveform using the window function generated in this way, and its spectrum distribution.

FIG. 3A shows a formant waveform when t _sk = 1 / 2t _win is set. in this case,
As the period of the periodic waveform of the period within formant formant center frequency f _f, it is generated in the pitch every period formant waveform example shown. FIG. 7B is a diagram showing the spectrum distribution of the formant waveform shown in FIG. 7A. The formant center frequency indicating the peak is _ff, and the frequency of the difference between the spectra is the pitch as shown. FIG. 4A shows a formant waveform when t _sk = 1 / 4t _win is set. As described above, the period of the formant center frequency f _f is the same as the period of the periodic waveform within formant waveforms, the formant waveform is generated every pitch period shown. FIG (b) is a diagram showing the spectral distribution of the formant waveforms shown in (a), the center frequency indicating the peak f _f
It is said. In this case, the skirt characteristic is controlled to be extended as compared with the spectrum distribution shown in FIG. 3, but the bandwidth near the peak is hardly changed.

FIG. 5A shows a formant waveform when t _sk = 1 / 8t _win is set. Period of the formant center frequency f _f is the same as the period of the periodic waveform within formant. This formant waveform is generated for each pitch cycle shown. FIG (b) is a diagram showing the spectral distribution of the formant waveforms shown in (a), the center frequency indicating the peak is set to f _f. In this case, FIG.
Although the skirt characteristic is controlled so as to be further extended as compared with the spectral distribution shown in (1), the bandwidth remains almost unchanged.

FIG. 6A shows a formant waveform when t _sk = 1 / 16t _win is set. Period of the formant center frequency f _f is the same as the period of the periodic waveform within formant waveforms, the formant waveform is generated every pitch period shown. FIG. 2B is a diagram showing a spectrum distribution of the formant waveform shown in FIG.
Center frequency indicating the peak is set to f _f. In this case, the skirt characteristic is controlled so as to be further extended as compared with the spectrum distribution shown in FIG. 5, but the bandwidth is still hardly changed.

FIG. 7A shows a formant waveform when t _sk = 1 / 32t _win is set. Period of the formant center frequency f _f is the same as the period of the periodic waveform within formant waveforms, the formant waveform is generated every pitch period shown. FIG. 2B is a diagram showing a spectrum distribution of the formant waveform shown in FIG.
Center frequency indicating the peak is set to f _f. In this case, the skirt characteristic is controlled so as to be further extended as compared with the spectrum distribution shown in FIG. 6, but the bandwidth is hardly changed.

FIG. 8A shows that t _sk = (t _win -1 / 4t)
The formant waveform when ( _win ) is set is shown.
In this case, the period of the formant center frequency f _f is equal to the period of the periodic waveform within formant waveforms, the formant waveform is generated every pitch period shown. FIG (b) is a diagram showing the spectral distribution of the formant waveforms shown in (a), the center frequency indicating the peak f _f
It is said. In this case, the spectrum distribution is similar to the spectrum distribution shown in FIG. Therefore, even when the time of the first half period portion of the window function is made longer than 1 / 2t _win as shown in this figure, the spectrum distribution as shown in FIGS. 4 to 7 is obtained.

According to the present invention, the window function is generated in this way, and the tone generator (T
FIG. 9 shows a block diagram of an electronic musical instrument having G). In this figure, a key-on (KON) signal and a signal such as a key code (KC) when a performance operator 1 such as a keyboard is operated, and a tone setting signal from a tone setting operator 2 are supplied to a controller 10. The KC data, key-on pulse (KONP) data, KON data, window function parameters (FWIN1,
FWIN2) and formant center frequency (FF) data are supplied to the TG 20. Further, the control unit 10
Also supplies the envelope parameter (EGPAR) to the TG 20. The TG 20 generates and outputs a formant sound according to the supplied data and parameters. The formant sound is emitted as a formant sound via a sound system including an amplifier and a speaker system (not shown).

Next, a circuit diagram of the tone generator (TG) 20 is shown in FIG. In this figure, key code (KC) data is supplied to a key code (KC) -F number (F _no ) conversion table 30 and is converted into F _no data. The _Fno data is input to one input terminal of the adder 31. The adder 31 adds the data input to the two input terminals and supplies an added output to one end of the AND gate 32. The addition data that has passed through the AND gate 32 is latched by a shift register 33 to which a sampling clock φ _s is supplied as a shift clock. By last addition data output from the shift register 33 is input to the other input terminal of the adder 31, so that the F _no data is accumulated. This F _no
Since the accumulated data of the data corresponds to the input key code, it corresponds to the phase of the pitch to be sounded, and the cumulative calculation power increases as the data is accumulated, but when the data overflows, Since it returns to the original state, it changes in a saw-tooth shape as shown in FIG.

The adder 31 and the AND gate 3
2. The accumulating means including the shift register 33 is reset to an initial value by a key-on pulse (KONP) shown in FIG. That is, the KONP is inverted and input to the AND gate 32 via the OR gate 37, whereby the AND gate 32 is closed by the KONP, and the data latched in the shift register 33 is all "0". It will be reset.

The window function parameters FWIN1, FW
IN2 is one of the parameters indicating the timbre of the formant sound, and is input to the selector 21 and one of the parameters is selected and output from the selector 21.
Window function parameter FWIN output from selector 21
1 or the window function parameter FWIN2 is input to one input terminal of the adder 22. The adder 22 adds the data input to the two input terminals and supplies the added data to one end of the AND gate 23, and outputs the carry output to the OR gate 24 to which the output from the AND gate 23 is supplied to one end. Is being entered. Thus, addition data passing through the AND gate 23 and OR gate 24 is latched by the shift register 25 to the sampling clock phi _s is supplied as a shift clock.
When the previous addition data output from the shift register 25 is input to the other input terminal of the adder 22, the window function parameter FWIN1 or the window function parameter FWIN2 is accumulated. Since the accumulated data is used as a read address of the window function waveform memory 27 described later, the accumulated data is used as phase data of the window function waveform.

Further, the MSB of the added data output from the AND gate 23 is supplied to the selector 21 as a selection signal. When the MSB is "0", the window function parameter FWIN1 is selected and output from the selector 21. , When this MSB is “1”, the window function parameter FWI
N2 is selected and output from the selector 21. The addition data output from the AND gate 23 is the accumulated data of the selected window function parameter FWIN1 or FWIN2, and indicates the phase of the window function waveform as described above. Is “0” during the first half period of the window function waveform.
The period during which B is “1” is the period of the second half cycle. Furthermore,
When the second half cycle ends and the adder 22 overflows, the carry output “1” is supplied to the OR gate 24, so that all “1” s are latched in the shift register 25, and the output from the shift register 25 is All will be fixed at "1".

Accordingly, the phase change speed of the first half cycle of the window function waveform is a speed based on the window function parameter FWIN1, and the phase change speed of the second half cycle of the window function waveform is a speed based on the window function parameter FWIN2. At the end of the cycle, the phase becomes fixed at the maximum phase. For this reason, for example, when the window function parameter FWIN1 is larger than the window function parameter FWIN2, the phase change is as shown in FIG. The accumulated data is supplied to the window function waveform memory 27 via the exclusive OR (EX-OR) circuit 26 as a read address. Therefore, the window function waveform is read from the window function waveform memory 27 according to the read address which is the accumulated data. The window function waveform memory 27
For example, a sample value of the waveform data of sin ² ωt is stored corresponding to the address.

Further, the accumulated data is supplied to one end of the EX-OR circuit 26, and the MSB of the accumulated data is supplied to the other end of the EX-OR circuit 26.
In the latter half period period when B becomes “1”, the read address which is the accumulated data is inverted and output from the EX-OR circuit 26. Therefore, the window function waveform memory 27 is read in the opposite direction during the latter half period, and the window function waveform is generated by reading the window function waveform memory 27 back and forth after all. . Therefore, it is sufficient that the window function waveform memory 27 stores a half cycle of the window function waveform.

Further, formant center frequency data FF, which is another parameter indicating the tone color of the formant sound, is input to one input terminal of the adder 38. The adder 38 adds the data input to the two input terminals and inputs the added data to one end of the AND gate 39. Addition data passing through the AND gate 39 is latched by the shift register 40 to the sampling clock phi _s is supplied as a shift clock. The previous addition data output from the shift register 40 is input to the other input terminal of the adder 38, whereby the formant center frequency data FF is accumulated. The accumulated data of this formant center frequency data FF is
This indicates the phase of the formant center frequency, and the accumulated data increases as the accumulation is performed, but returns to its original state when it overflows, so that it periodically changes in a sawtooth shape as shown in FIG. . The accumulated data indicating this phase is supplied to the waveform memory 41 as a read address, and the formant center frequency signal is periodically read from the waveform memory 41. The waveform memory 41 has, for example, 1
A periodic sine waveform sinωt is stored. Of course, a waveform other than a sine waveform may be stored to obtain a more complex formant waveform.

The MSB of the accumulated data indicating the pitch phase from the shift register 33 is, for example, the delay means 3 for delaying data by one sampling clock φ _s.
4 and to the inverter 35. Further, the outputs of the delay means 34 and the inverter 35 are applied to the AND gate 36, so that the AND gate 3
From FIG. 6, a differentiated pulse shown in FIG. 11C in which the falling edge of the MSB is differentiated can be obtained. The differentiated pulse is passed through the OR gate 37 as a reset pulse to the AND gate 32, AND gate 23, AND gate 3
9 is applied. The reset pulse inverted and input to the AND gate 32 is applied to the AND gate 32.
Is closed, the latch data of the shift register 33 that latches the output of the AND gate 32 is all "0". Thus, the accumulating means including the adder 31, the AND gate 32, and the shift register 33 is reset every pitch period.

The reset pulse inverted and input to the AND gate 23 closes the AND gate 23, so that the latch data of the shift register 25 that latches the output of the AND gate 23 is all "0". Thereby, the adder 32, the AND gate 23, and the OR gate 2
4, the accumulating means including the shift register 25 is reset every pitch period. Further, the reset pulse inverted and input to the AND gate 39 is applied to the AND gate 3
9 is closed, the latch data of the shift register 40 that latches the output of the AND gate 39 is all "0". Thus, the accumulating means including the adder 38, the AND gate 39, and the shift register 40 is reset every pitch period.

Thus, at the starting point of the pitch cycle,
The phase of the pitch, the phase of the window function waveform, and the phase of the formant center frequency are all reset to the initial value, and each waveform is always generated from the same phase. For this reason, the spectrum distribution of the formant sound is always kept constant and is prevented from fluctuating. Further, since the phase of the window function waveform is reset to the initial value for each pitch, the window function waveform is generated for each pitch. The formant center frequency signal generated from the waveform memory 41 and the window function waveform generated from the window function waveform memory 27 are input to the multiplier 28 and multiplied by the two signals. The formant waveform shown in FIG. 11 (G) is generated. Since the window function is generated for each pitch as described above, this formant waveform is generated for each pitch as illustrated. This formant waveform is further multiplied by the envelope signal by the multiplier 29 and output as a formant sound. This envelope signal indicates a key depressed state by K
It is generated by an envelope generator 42 to which an EGPAR corresponding to an ON signal and key touch information is supplied.

The formant waveform can be controlled as shown in FIGS. 3 to 8 by changing the window function parameter FWIN1 for determining the phase velocity of the first half cycle. In this case, the window function parameter FWIN2
The window function parameter F is set so that the time width t _win of the window function is not changed even if the window function parameter FWIN1 is changed.
WIN2 and the window function parameter FWIN1 have a constant value. Alternatively, the window function parameter FWIN
The window function parameter FWIN2 may be inverted by inverting 1. Further, the window function parameters FWIN1 and FWIN2 can continuously change parameter values.

In the formant type sound source described above, the formant center frequency data FF and the window function parameter F are set according to the operation of the tone setting operator 2 shown in FIG.
WIN1 and a window function parameter FWIN2 are set. Usually, the window function parameters FWIN1 and FWIN2 are set so that the time width of the window function does not exceed the pitch cycle. However, multiple formant sound sources, or
For example, by providing two systems in parallel and generating a formant waveform sequentially or alternately according to the pitch cycle, a formant waveform having a time width exceeding the pitch cycle can also be generated. Further, the formant center frequency data FF, the window function parameter FWIN1, and the window function parameter FWIN2 may be set in accordance with the pitch and the sound range to be generated. Further, assuming that the sample value data of the waveform memories 27 and 41 is logarithmic data, the multipliers 28 and 29
Can be replaced with an adder with a simple configuration. However, it is necessary to convert the logarithm to an antilog by the log / antilog conversion means before outputting.

Further, two or more tone generators TG are arranged in parallel as shown in FIG. 12, and formant waveforms having different formant center frequencies are generated from each of them, and these are synthesized and output by the adder 20-4. You may do so. In this case, KC data, KON
P, KON data are shared by each TG 20-1, 20-2,
20-3 and the window function parameter FWIN
11, window function parameter FWIN21, formant center frequency FF1, and envelope parameter EGPA
R1 is supplied to the TG 20-1, and the window function parameter FWI
N12, window function parameter FWIN22, formant center frequency FF2, and envelope parameter EGP
AR2 is supplied to the TG 20-2, and the window function parameter FW
IN13, window function parameter FWIN23, formant center frequency FF3, and envelope parameter EG
PAR3 is supplied to TG 20-3.

Then, for example, from TG 20-1 to FIG.
13A, a formant waveform having the spectrum distribution characteristic shown in FIG. 13B is generated from the TG 20-2, and a formant waveform having the spectrum distribution characteristic shown in FIG. 13C is generated from the TG 20-3. become. The formant waveforms having different spectral distribution characteristics are added to the adder 20-4.
By synthesizing and outputting, it is possible to obtain more various formant sounds.

[0038]

Since the present invention is configured as described above, the skirt characteristic of the formant waveform can be changed without changing the bandwidth by changing the first half period time of the window function with respect to the time width. It can be varied and the hardware for this control can be simplified. Further, since the first half cycle time with respect to the time width of the window function can be continuously changed, the skirt characteristic can be finely controlled. Further, if a window function based on a sine function is used, the convergence of the window function is fast, so that it is possible to prevent the formants from being superimposed. Further, when a formant sound is generated by synthesizing a plurality of formant waveforms, it is possible to generate a more varied formant sound.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating the principle of generating a window function according to the present invention.

FIG. 2 is a diagram showing a window function when t _sk is changed.

FIG. 3 is a diagram showing a formant waveform and its spectrum when t _sk = １ ／ t _win .

4 is a diagram showing formant waveform and its spectrum in the case of t _sk = 1 / 4t _win.

5 is a diagram showing formant waveform and its spectrum in the case of t _sk = 1 / 8t _win.

FIG. 6 is a diagram showing a formant waveform and its spectrum when t _sk = 1 / 16t _win .

FIG. 7 is a diagram showing a formant waveform and its spectrum when t _sk = 1 / 32t _win .

FIG. 8 is a diagram showing a formant waveform and its spectrum when t _sk = (t _win −１ ／ t _win ).

FIG. 9 is a block diagram of an electronic musical instrument having a formant type sound source according to the present invention.

FIG. 10 is a block diagram of a formant-type sound source according to the present invention.

FIG. 11 is a waveform diagram of each part of the sound source of the formant method of the present invention.

FIG. 12 is a diagram showing a modification of the sound source of the formant system of the present invention.

FIG. 13 is a diagram showing a spectrum in a modification of the sound source of the formant system of the present invention.

FIG. 14 is a diagram illustrating the principle of forming a formant.

FIG. 15 is a block diagram of a conventional formant type sound source.

FIG. 16 is a diagram showing waveforms at various parts of a conventional sound source of the formant method.

FIG. 17 is a diagram showing a spectrum of a formant generated from a conventional sound source of the formant method.

FIG. 18 is a diagram showing a spectrum of another formant generated from a sound source of a conventional formant method.

FIG. 19 is a diagram showing a spectrum of still another formant generated from a sound source of a conventional formant method.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 Performance operator 2 Tone setting operator 10 Control part 20,20-1,20-2,20-3 Tone generator 21,131,134 Selector 20-4,22,31,38,104,105 Adder 23 , 26, 32, 39 AND gate 24, 37 OR gate 25, 33, 40 Shift register 26 EX-OR 27 Window function waveform memory 28, 29 Multiplier 30 KC-F _no converter 34 Delay circuit 35 Inverter 41 Waveform memory 42 Envelope generator 106 Log / antilog converter 107 DAC 111, 121, 132 Phase accumulator 112 Differentiator 122 Sine wave table 133 Sine square wave table 135 Bit shifter

──────────────────────────────────────────────────続 き Continued on the front page (58) Field surveyed (Int.Cl. ⁷ , DB name) G10L 13/00 G10H 7/08

Claims (1)

(57) [Claims] 1. A window function time width generating means for setting a time width of a window function and a first half period time width of a window function in accordance with each independent tone color information, and a window output from the window function time width generating means. depending on the function information and the pitch information, and a window function generating means for generating a window function signal for each of the pitch period with the set time width and the former half period duration, formant center frequency in response to said tone color information A periodic function generating means for generating a signal, and the window function signal generated from the window function generating means is used to generate a signal from the periodic waveform generated from the periodic function generating means.
A formant sound source characterized in that a formant sound is generated by amplitude-modulating the formant center frequency signal.