JPH02271396A - Formant sound synthesizing device - Google Patents

Abstract

PURPOSE:To synthesize diverse timbre and high-pitch formant sound by multiplying a formant sound pitch by the period waveform of independent desired timbre and a necessary window waveform, and further multiplying the result by a necessary damp waveform. CONSTITUTION:The window waveform (d) of a modulated wave is generated by the modulated wave generating circuit 3 consisting of a phase accumulator 31, applied with formant band width information BW controlled with a period waveform (b) corresponding to formant sound pitch outputted by a pulse generating circuit such as an accumulator 11 applied with the fundamental frequency information Fo on the fundamental pitch of formant sound pitch, and a sine square or its power circuit 32. A multiplier 5 multiplies this waveform (d) by the period waveform (e) corresponding to the desired timbre of a carrier generating circuit 2 such as an accumulator 21 controlled with a formant center frequency Ff and a multiplier 6 further multiplies the result with an output waveform (g) from a forcible damp waveform memory 4 to easily synthesize a high-pitch formant sound with diverse timbre that a window does not overlap with even when the period of the waveform (b) is shorter than the waveform (d).

Description

[Detailed Description of the Invention] [Industrial Field of Application] The present invention relates to a device for electrically synthesizing ormantic sounds (such as vowels of human voices and sounds of natural instruments), and particularly relates to a device for electrically synthesizing ormantic sounds (such as vowels of human voices and sounds of natural instruments), and particularly, The present invention relates to a firmant sound synthesis device capable of synthesizing various formant sounds.

[Prior art] Fig. 11 shows an amplitude waveform diagram (a) and an amplitude spectrum diagram (b) when a bass singer utters the "A' sound at pitch C5.
shows. Further, FIG. 12 shows a general amplitude spectrum distribution curve of such a formant sound. In FIG. 12, ff is the formant center frequency and Bw is the formant bandwidth.

Such a formant sound waveform has the 13th frequency ft.
The first waveform as shown in Figure (b) and the first waveform with a longer period of 1/f rise from the reference level to a high level and then fall towards the reference level (Figure 13(a) )
A second waveform (hereinafter referred to as a window waveform) as shown in (hereinafter referred to as a window waveform) is repeatedly generated while initially setting the frequency f0 corresponding to the pitch (fundamental pitch) of the formant sound to be generated, and these waveforms are It can be obtained by multiplying. As a specific configuration of such a device, an example in which it is used as a sound source for an electronic musical instrument is disclosed in Japanese Patent Publication No. 59-19352 by the present applicant, for example.

The electronic musical instrument disclosed in Japanese Patent Publication No. 59-19352 changes the timbre by varying the rate of change of the address signal in the first and second waveform generating means, that is, the period of the first and second waveforms. It can be changed arbitrarily. However, since the basic readout waveform is constant, there is a limit to the range of timbre change simply by making the readout period variable.

Also, Computer Music Journal 8 (3
) On pages 9 to 14, a method using a window waveform obtained by transforming a sine wave by calculating S (k) = Ge-aksin (+φ for ω) is proposed as an FOF method for IRCAM. In this method, by varying α and ω as parameters, it is possible to obtain a variety of window waveforms, thereby making it possible to vary the formant bandwidth Bw and, therefore, the timbre more broadly as shown in Fig. 14. .

However, in this FOF method, since the window waveform is basically an exponentially decaying waveform, the decay time is long and phonemes often distort. Therefore, it is necessary to provide as many phoneme generators (in particular, the second waveform generating means and the multiplication means) in parallel, resulting in an inconvenience that the scale of the hardware configuration increases.

Therefore, the inventor of the present invention created the window waveform as follows: 5(t)zsin2nkt (however, n! 2 ”
). This waveform is kt-
Since it is completely attenuated up to 2π, the attenuation time is short and the number of phoneme overlaps is reduced, making it possible to reduce the number of generators (phoneme generators or operators). In addition, two or more generators are used to alternately generate phonemes, and the structure is configured so that it can cope with the occurrence of overlapping phonemes equal to the number of generators, making it possible to synthesize formant sounds up to even higher pitches (hereinafter referred to as , called the advance method).

As described above, in order to generate a formant sound waveform, a 'shaped window corresponding to the timbre is applied to the first waveform of the formant frequency, for example, a sine wave. This window occurs for each basic pitch. The phase of the sine wave is reset at each starting point of the window. But then,
When producing high-pitched sounds, the window generation period must be shortened. On the other hand, the formant band width Bw has a property that the wider the window width W (see FIG. 14(a)), the narrower it becomes. Therefore, in order to keep the formant bandwidth constant regardless of the pitch, it is necessary to keep the window width W constant regardless of the pitch. When synthesizing low pitch sounds, the windows do not overlap, so a single generator is fine. However, as the pitch becomes higher and the windows overlap as shown in FIG. 14(b), it becomes a problem how to process the overlapping portions of the windows with a single generator.

Even in the above-described prior method by the present inventor, since the number of generators used is finite, there is always a limit to the pitch that can be synthesized, that is, a minimum pitch at which the window waveforms of each generator do not overlap. According to the inventor's experiments, in the case of the two-gear generator system, f, wlkH
It was about z.

[Problems to be Solved by the Invention] An object of the present invention is to provide a formant sound synthesis device that has a simple configuration, can generate various tones, and can synthesize higher pitch formant sounds.

[Means and effects for solving the problem] In order to achieve this object, the present invention provides a formant sound independent of the pitch of the formant sound to be synthesized, which is repeatedly initialized and generated at a period corresponding to the pitch. A first waveform having a repeating period and a second waveform (window waveform) having a longer period than the first waveform are repeatedly initialized at the basic pitch of the formant sound, and these first and second waveforms are generated by repeatedly initializing them at the basic pitch of the formant sound. In a formant sound synthesis device that synthesizes a formant sound signal by multiplying two waveforms, a waveform of the square of a sine wave or a power thereof is used as the second waveform, and at least the basic pitch is the same as that of the second waveform. When it is shorter than the cycle, generate a third waveform starting from 1 and smoothly reaching O in synchronization with the first and second waveform generation operations, and multiplying this with the first and second waveforms. It is characterized by

[Operation] In this invention, to explain by giving a specific example, the first
As waveforms 3 to 3, waveforms as shown in FIGS. 1(A) to 1(C) are generated, respectively. In FIG. 1, (A) is a sine wave (first waveform) having a frequency equal to the formant center frequency ft. (B) and (Bo) are 1
/ f A window envelope (
(2nd waveform), (B) shows that the period W of the window is period 1
/f0 is the waveform when it is shorter, (Bo) is the waveform when it is longer, (C) is the forced dump envelope (third waveform) that is a feature of this invention, and here, cos''
"An example using the first 1/4 waveform of 2πfot is shown. Formant sound is synthesized by multiplying these three waveforms. (D) is (A), (B) and (C) The formant sound waveform obtained by multiplying the waveform shown in is shown, and (Do) is (A), (Bo) and (C)
The formant sound waveform obtained by multiplying the waveform shown in is shown.

That is, in this invention, when attempting to synthesize a formant sound with a pitch shorter than the window period W, the window waveforms are forcibly reset as shown in FIG. 1 (Bo), thereby preventing the overlap of the window waveforms. Instead, the formant sounds are synthesized using a single generator. However, if this forced reset is continued, the spectrum of the formant sound, that is, the timbre, will be significantly different from the long pitch sound. Therefore,
In this invention, as shown in (C), the forced dump waveform starts from 1 and smoothly reaches 0 at the period of the basic pitch and in synchronization with the first and second waveform generation operations. 1 and N2 waveforms. As a result, the operation of the formant sound synthesizer is to convert the smoothly attenuating window waveform obtained by multiplying the window waveform (Bo) by the forced dump waveform (C) into the sine wave ( This is equivalent to multiplying by A). Therefore, changes in formant spectrum (timbre) due to pitch changes can be reduced.

In addition, as the forced dump waveform (C), use the J waveform that starts from 1 and continues to O. Here, in particular, 1/
Forced dump waveform (C) regardless of the size of fo and W
In the case of using a method of multiplying , it is desirable to have a waveform in which (1) the flat portion is appropriately long, and (2) the slope dy/dx of the starting point and ending point is O.

The reason for this is that when the phonemes are shorter than 1/f0, that is, when they do not overlap, they do not affect the original phoneme envelope, and when the forced dump waveform becomes discontinuous, there is an unnecessary This is to prevent sideband waves from occurring.

[Example] Hereinafter, this invention will be explained in detail based on examples.

FIG. 2 is a block diagram showing the configuration of a formant sound synthesis device according to an embodiment of the present invention.

Moreover, FIGS. 3-3 are operation timing waveform charts of each part of the apparatus of FIG. 2.

The device shown in the figure includes a pulse generation circuit 1, a carrier waveform generation circuit 2
, a window waveform generation circuit 3, a forced dump waveform generation circuit 4, multipliers 5, 6, 7, a D/A converter 8, and the like. Such a device is used, for example, as a sound source for an electronic musical instrument, and is provided with a formant center frequency information value Ffs, a formant fundamental frequency information value FO, and a formant bandwidth value BI1% given from a sound generation control means (not shown).
A firmant sound is synthesized based on the formant sound amplitude (envelope) waveform data ENV, etc.

In the pulse generation circuit 1, a 0-phase accumulator 11 consisting of a phase accumulator 11 and a differentiator 12 generates the fillant fundamental frequency information value F in synchronization with a predetermined clock pulse φ. Accumulate.

Here, the formant fundamental frequency information value F0 is set to a value corresponding to the fundamental frequency of the formant sound to be generated, that is, the pitch (basic pitch), and the phase accumulator 11 has an accumulated value QFO ((1 =1.2,3
．． ......) is output.

Figure 3 ■ shows the accumulated value output of the phase accumulator 11, and Figure 3 ■'' shows the most significant bit (MSB) of this accumulated value.
shows the output of Differentiator 12 differentiates the falling edge of this MSB output. Therefore, this pulse generation circuit 1 generates a pulse signal (see Fig. 3) as the output of this differentiator 12 with a period corresponding to the pitch of the formant sound to be generated.
occurs repeatedly. This pulse signal is supplied to carrier waveform generation circuit 2 and window waveform generation circuit 3 as an initial setting signal. Further, the accumulated value output QFo of the phase accumulator 11 is supplied as a forced dump river waveform mesori 4 head address signal.

In the carrier waveform generation circuit 2, a 0-phase accumulator 21 consisting of a phase accumulator 21 and a sine wave memory 22 accumulates the formant center frequency information value Fr in synchronization with the clock pulse φ, and calculates the accumulated value qFr (Q-
1, 2, 3, . . . ) are sequentially output as read address signals of the sine wave memory 22. The sine wave memory 22 stores the sequential sample point amplitude values sin θ of one period of the sine wave at each address, and the amplitude values sinθ are stored at the addresses designated by the address signal (accumulated value qFr) output from the accumulator 21. The sine wave amplitude values that are present are sequentially read out. As a result, the sine wave memory 22 sequentially outputs sample point amplitude values sIn 2πfπ of the sine wave at a frequency ff corresponding to the frequency information value F in accordance with the clock pulse φ.

Here, formant center frequency information value F.

is the pitch f of the formant sound to be generated. It is set to a value corresponding to the formant center frequency f, which is one of the parameters that independently indicates the timbre of the formant sound. Therefore, this carrier waveform generating circuit 1 generates a sine wave (see FIG. ■) will occur.

The window waveform generation circuit 3 includes a phase accumulator 31
and a 0-phase accumulator 31 consisting of a sine square wave memory 32 converts the bandwidth value By into the clock pulse φ
The accumulated value QBw (q=f, 2
,3. ...) are sequentially output as read address signals of the sine square wave memory 32. Moreover, as shown in FIG. 3, this phase accumulator 31 has an accumulated value of 2π.
When the value corresponding to 2π is reached, the accumulation operation is stopped and the value corresponding to 2π is continued to be output. In the sine square wave memory 32, sequential sample point amplitude values of 1/2 cycle of the sine square wave sin'θ are stored at each address, and the accumulator 3! The sine square wave amplitude value 5in stored at the address specified by the address signal (cumulative value qBw) output from
2 kt are sequentially read out (Fig. 3).

The waveform memory 4k for forced dump is 1 of (1-sin”θ)
Sequential sampling point amplitude values of /4 cycles are stored in each address, and the address x / 4k memory is specified based on the address signal (accumulated values qF, mx) output from the accumulator 11 of the pulse generation circuit 1. The forced dump waveform amplitude values that have been set are sequentially read out. As a result, the frequency information value F is output from the forced dump waveform memory 4 according to the clock pulse φ. Sequential sample point amplitude value 1-5in2nπfot of forced dump waveform of frequency f0 corresponding to
/2 are sequentially output (■ in Figure 3).

In the multiplier 5, the sine wave amplitude value sin 2πf,t (Fig. 3) output from the carrier waveform generation circuit 2 and the sine square wave amplitude value sIn''kt (3 Figure ■).Thereby, as the output of multiplier 5, (sin 2ytfy t)
The product of (sln''k t ) (Fig. 3) is obtained.

Multiplier 6 outputs the product (sin2 r
The forced dump waveform (1-s1021πf, t
/2). As a result, the output of the multiplier 6 is the first waveform (sin 2 πf, t), the second waveform (sin"k t), and the third waveform (1-5in").
π fot/2) (Fig. 3 ■) is obtained. This product indicates the sequential sample point amplitude value of the formant signal.

The multiplier 7 is for adding an envelope to the formant signal, and the output of the multiplier 5 (sin 2 y
r f y t)-(5ln2k t)・(1-5in
”πfOt/2) and the above envelope waveform data ENV
Multiply by As a result, the output of the multiplier 7 is the product of the formant sound amplitude (1 and the envelope waveform value ENV) (sin 2πfπ)・(sin”kt)・(1
−sin2nπfot/2)4NV is obtained. This product indicates the sequential sample point amplitude value of the formant sound.

The DA converter 8 converts the sequential sample point amplitude value data of the formant sound output from the multiplier 7 into an analog signal.

This analog signal is converted into sound through a sound system including an amplifier and a speaker system (not shown), and is produced as formant sound.

Note that FIG. 4 shows the forced dump waveform output from the forced dump waveform memory 4. This waveform is expressed as 1 + (5l
n2i )'(4k+2)π≦x< (4k+4) When π, 1−(
・i・′ (Issue-Issue).

In the figure, the parameter n is set to 1, respectively.
, 4.Curve 1-( when set to 16, 64, 256
sin'tusk)' wo show.

Next, using the apparatus shown in Fig. 2, the parameter n and window width W are set variously so that the formant center frequency becomes f.
An example will be described in which a formant sound is synthesized at t = 4 kHz and the fundamental frequency is f = 200 Hz.

Figures 5 to 7 show the window width W and basic pitch 1/f0.
are set equal, and the parameters n are set to 0, 16, .
The waveform (a) and amplitude spectrum (b) of a formant sound are shown in the case where n=0 means no forced dumping.

By comparing these FIGS. 5 to 7, it can be seen that if the window width W is shorter than the basic pitch 1/f0, the effect of multiplication by the forced dump envelope waveform will hardly appear.

In FIGS. 8 to 10, the basic pitch I/fo is set to 1/2 of the window width W, and the parameter port is set to O (no forced dump) as above.

Waveform of formant sound when setting 16.64k (a
) and amplitude spectrum (b), these 8th to
By comparing FIG. 10, when the window width W is longer than the basic pitch 1/fo, a good formant with a narrow band width can be obtained by multiplying by the forced dump waveform.

In addition, in the above-mentioned embodiment, an example was shown in which a sine square wave was used as the window waveform, but it is also possible to use a sine square wave obtained by raising this to a power. In this case, by changing S, the formant bandwidth By can be adjusted. Furthermore, by making S different between the first half and the second half of the window waveform to make the window waveform asymmetric between the first half and the second half, it is possible to adjust the balance between the formant band width Bw and the bottom waveform. Furthermore, by setting S in the latter half to a large value, it is possible to make it less susceptible to the influence of forced dumping.

In addition, in the above description, a sine wave is used as the carrier waveform in order to synthesize a formant sound waveform with a unimodal spectrum, but by using a signal having frequency components corresponding to each peak of the amplitude spectrum as the carrier waveform, , it is possible to synthesize a formant sound waveform that has multiple peaks in its amplitude spectrum.

Further, in the above description, multiplication is performed using a multiplier, but by handling each data logarithmically, an adder can be used instead of the multiplier. In this case, a log/antilog converter is ultimately required, but the calculation speed can be increased.

[Effects of the Invention] As explained above, conventionally, each phoneme is synthesized using two or more operators and then added by an adder, so that the width W of the window waveform is long (the formant bandwidth is However, according to this invention, each time a new phoneme begins, if the previous phoneme continues, it is forcibly dumped.
One formant sound can be synthesized with a single operator. Therefore, although a waveform generator for forced dumping is required, the number of operators can be reduced, and as a result, the hardware configuration can be simplified. Alternatively, a larger number of formant sounds can be generated simultaneously with the same hardware configuration. In this case, the basic pitch is 17f.

The longer the window width W becomes, the stronger the influence of forced dump appears, and the more complex control logic such as generating phonemes at a cycle of f0/n (n is the number of operators) and accumulating them, becomes more pronounced. This is no longer necessary, and greater benefits can be obtained, such as an increase in the number of firmant sounds that can be generated simultaneously.

Also, even if configured to always force dump,
When the window width W of a phoneme is short, it is possible to reduce the influence of forced dumping and to obtain a performance comparable to that of the conventional system.

Furthermore, in the conventional method, phonemes with a period longer than f, /n cannot be handled. Such phonemes often occur when the formant width is narrowed in the high range, but according to the present invention, formant sounds can be synthesized without being subject to such limitations.

[Brief explanation of drawings]

FIG. 1 is a waveform diagram for explaining the basic operation of the present invention, FIG. 2 is a block diagram of the formant sound synthesis device according to the present invention, and FIG. 3 is the operation timing of each part of the device shown in FIG. 2. Figure 4 is a diagram of the forced dump envelope waveform generated by the device in Figure 2. Figures 5 to 10 are formant sounds synthesized when various parameters are set for the device in Figure 3. In each figure, (a) is a waveform diagram, (b) is an amplitude spectrum diagram, Figure 11 (a) and (b) are a waveform diagram and amplitude spectrum diagram showing an example of natural fillant sound, and Figure 12 is 13(a) and 13(b) are diagrams showing carrier waveforms and window waveforms used in a conventional formant synthesis device. FIGS. 14(a) and 14(b) are schematic spectral diagrams of natural formant sounds. FIG. 4 is a window waveform diagram for explaining the operation of a method previously attempted by the present inventor. Pulse generation circuit carrier waveform generation circuit modulation (window) waveform generation circuit forced dump waveform memory, 6: Multiplier: Sine square wave memory procedure amendment (method) August 15, 1989 Commissioner of the Japan Patent Office Yoshi 1) Fumi Mata 1. Indication of the case 1999 Patent Application No. 91761 2. Name of the invention Formant sound synthesizer 3. Person making the amendment Relationship to the case Patent applicant address 10-1 Nakazawa-cho, Hamamatsu City, Shizuoka Prefecture Name ( 407
) Yamaha Corporation Representative Hiroshi Kawakami 4, Agent 105 Address 2-8-1 Toranomon, Minato-ku, Tokyo June 30, 1999 (July 25, 1999) 6, Subject to amendment

Claims (2)

[Claims] (1) A first waveform generating means that generates a first waveform having a predetermined first repetition period that is arbitrarily set corresponding to a desired timbre independently of the pitch of the formant sound to be synthesized. It consists of the square of a sine wave or its power waveform, and the first waveform is
a second waveform generating means for generating a second waveform having a second period longer than the repetition period of the formant sound; and a pulse signal generating means for repeatedly generating a pulse signal having a period corresponding to the pitch of the formant sound; The waveform generation operation in the first and second waveform generation means is repeatedly initialized by the pulse signal, and the first and second waveform generation means are
a control means for repeatedly generating a waveform of , when at least the cycle of the pulse signal is shorter than the second cycle, a third waveform that starts from 1 and smoothly reaches 0 in synchronization with the pulse signal at the cycle of the pulse signal; a third waveform generation means for generating a waveform; and a multiplication means for multiplying the waveforms generated from the first to third waveform generation means, and generates a formant sound waveform as an output of the multiplication means. A formant sound synthesis device characterized by: (2) The third waveform is expressed as 1-(sin^2(X/4)^n (4k+2)π≦x< when 4kπ≦x<(4k+2)π, where x is the phase and k is a positive integer. When (4k+4)π, 1−{si
2. The formant sound synthesis device according to claim 1, wherein the formant sound synthesizer has a waveform represented by n^2([π/2]/[X/4])}^n.