JP2504178B2 - Formant sound synthesizer - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a device for electrically synthesizing formant sounds such as vowels of human voices and natural musical instrument sounds, and more particularly, to a device having a simpler structure and more diverse structure. The present invention relates to a formant sound synthesizer capable of synthesizing different formant sounds.

[Prior art] Figure 11 shows views amplitude waveform when the bus singer uttered the "a" sound pitch C ₃ (a) and the amplitude spectrum diagram of (b). Further, FIG. 12 shows a general amplitude spectrum distribution curve of such a formant sound. In FIG. 12, f _f is the formant center frequency, and B _W is the formant bandwidth.

Such a formant sound waveform has a first waveform of frequency f _f as shown in FIG. 13 (b) and a period 1 / of the first waveform.
A second waveform (hereinafter referred to as a window waveform) as shown in FIG. 13 (a) that rises from the reference level toward a high level at a cycle longer than f _{f and} then falls toward the reference level,
Pitch of the formant sound to be generated (basic pitch)
Occurs repeatedly with initial setting at the frequency f _o corresponding to
And it can be obtained by multiplying these waveforms. As a specific configuration of such an apparatus, for example, Japanese Patent Publication No. 59-19352 by the present applicant discloses an example in which it is used as a sound source of an electronic musical instrument.

The electronic musical instrument disclosed in Japanese Examined Patent Publication No. 59-19352 has a relatively timbre by varying the changing speed 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,
Only by making the reading cycle variable, there was a limit to the change range of the timbre.

Also, Computer Music Journal 8
(3) On pages 9 to 14, as a FOF method of IRCAM, a method of using a window waveform obtained by ^modifying a sine wave by an operation of S (k) = Ge- ^αk sin (ωk + φ) is proposed. In this method, a variety of window waveforms can be obtained by varying α and ω as parameters, and as a result, the formant band width B _W as shown in FIG. 14, and thus the timbre, can be changed in a wider range. is there.

However, in this FOF method, since the window waveform is basically an exponentially decaying waveform, the decay time is long and a lot of phoneme overlap occurs. For this reason, it is necessary to provide as many phoneme generators (especially the second waveform generating means and the multiplying means) as the number of overlaps in parallel, which disadvantageously increases the scale of the hardware configuration.

Therefore, the present inventor has tried to use a waveform of S (t) = sin ²ⁿ kt (where n = 2 ^sa ) as the window waveform. This waveform is kt = 2π
Since it is completely attenuated by, the decay time is short, the phoneme overlap is small, and the number of generators (phoneme generators or operators) can be reduced. In addition, two or more generators are used to alternately generate phonemes, and it is configured so as to be able to cope with the occurrence of phoneme overlaps by the number of generators, and formant sounds can be synthesized up to a higher pitch (below. , Called the preceding method).

As described above, in order to generate the formant sound waveform, the first waveform of the formant frequency, for example, a sine wave, is provided with a window having a shape corresponding to the tone color. This window is generated for each basic pitch. The phase of the sine wave is
Reset at every start of window. Therefore, when producing a sound with a high pitch, the generation period of the window must be shortened. On the other hand, the formant band width B _W has a property that it becomes narrower as the window width W (see FIG. 14 (a)) becomes wider. Therefore, in order to make the formant band width constant regardless of the pitch, it is necessary to keep the width W of the window constant regardless of the pitch. When synthesizing low pitch sounds, the windows do not overlap, so there is no problem with a single generator. However, when the pitch becomes high and the windows overlap as shown in FIG. 14 (b), how to handle the overlapping parts of the windows with a single generator becomes a problem.

Since the number of generators used is also finite in the above prior art by the present inventor, there is always a limit of synthesizable pitch, that is, a minimum pitch in which window waveforms do not overlap in each generator. According to the experiments by the inventor, in the case of the two-generator system, f _o = 1 kHz.

[Problems to be Solved by the Invention] An object of the present invention is to provide a formant sound synthesizing apparatus which has a simple structure and is capable of generating various timbres and capable of synthesizing formant sounds of higher pitch.

[Means and Actions for Solving the Problem] In order to achieve this object, the present invention is independent of the pitch of the formant sound to be synthesized, which is repeatedly initialized and generated at a cycle corresponding to the pitch. A first waveform having a repeating period and a second waveform (window waveform) having a period longer than the first waveform are generated while repeatedly initializing at the basic pitch of the formant tone.
In a formant sound synthesizer for synthesizing a formant sound signal by multiplying a second waveform with a second waveform, a sine wave squared or a power thereof is used as the second waveform, and at least the basic pitch is the second waveform. When the period is shorter than the waveform period, a third waveform starting from 1 and smoothly reaching 0 is generated in synchronization with the first and second waveform generating operations, and this is multiplied with the first and second waveforms. It is characterized by doing.

[Operation] In the present invention, a concrete example will be described. Waveforms as shown in FIGS. 1A to 1C are generated as the first to third waveforms, respectively. In FIG. 1, (A) is a sine wave (first waveform) having a frequency equal to the formant center frequency f _f . (B) and (B ') are 1 / f _f
Made from longer period of the sine wave of the square or a power waveform in the window for the envelope to be repeatedly generated in the period of 1 / f _o corresponding to the basic pitch (second waveform),
(B) shows the waveform when the window period W is shorter than the period 1 / f _o , and (B ') shows the waveform when it is long. (C) is an envelope for forced dumping (third waveform), which is a feature of the present invention. Here, an example using the first 1/4 waveform of cos ^{1 / m 2} πf _o t is shown. Then, the formant sound is synthesized by multiplying these three waveforms.
(D) shows a formant sound waveform obtained by multiplying the waveforms shown in (A), (B) and (C), and (D ') shows the waveforms shown in (A), (B') and (C). The formant sound waveform obtained by multiplying by is shown.

That is, in the present invention, the window period W
When trying to synthesize a formant sound with a shorter pitch, the window waveforms are forcibly reset as shown in FIG. 1 (B ') to eliminate the overlapping of the window waveforms, and formant sounds are generated by a single generator. I try to synthesize them. However, if the reset is forcibly reset in this way, the spectrum of the formant sound, that is, the timbre, is significantly different from that of the long-pitch sound. Therefore, in the present invention, as shown in (C), a forced dump waveform having a basic pitch period and starting from 1 and smoothly connecting to 0 in synchronization with the first and second waveform generating operations, Multiply with the first and second waveforms. As a result, in the operation of the formant sound synthesizer, a smoothly decaying window waveform obtained by multiplying the window waveform (B ') and the forced dump waveform (C) is a sine wave (the first waveform). It is equivalent to multiplication with A). Therefore, it is possible to reduce the change in the formant spectrum (timbre) due to the change in pitch.

As the forced dump waveform (C), a waveform “starting from 1 and connecting to 0” is used. Where 1 /
When the method for multiplying the force dampening waveform (C) regardless of the size of the f _o and W are flat portion appropriately long waveform is desired slope dy / dx of the start and end points is 0 .

The reason is that when the phonemes are shorter than 1 / f _o , 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 extra amount when this waveform is multiplied. This is to prevent such sidebands from occurring.

[Examples] Hereinafter, the present invention will be described in detail based on examples.

FIG. 2 is a block diagram showing the configuration of a formant sound synthesizer according to an embodiment of the present invention. 3 to 3 are operation timing waveform charts of respective parts of the apparatus shown in FIG.

The apparatus shown in FIG. 1 includes a pulse generating circuit 1, a carrier waveform generating circuit 2, a window waveform generating circuit 3, a forced dump waveform generating circuit 4, multipliers 5, 6, 7 and a D / A converter 8. . Such a device is used, for example, as a sound source of an electronic musical instrument, and is provided with a formant center frequency information value F _f , a formant fundamental frequency information value F _o , a formant band width value B _W , and a formant sound of a formant sound, which are given from a sound generation control means (not shown). Formant sound is synthesized based on amplitude (envelope) waveform data ENV and the like.

The pulse generation circuit 1 includes a phase accumulator 11 and a differentiator 12. The phase accumulator 11 accumulates the formant fundamental frequency information value F _o in synchronization with a predetermined clock pulse φ. Here, the formant fundamental frequency information value F _o 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 calculates the accumulated value qF _o ( q =
1,2,3, ...) are output. FIG. 3 shows the accumulated value output of the phase accumulator 11, and FIG. 3'shows the output of the most significant bit (MSB) of this accumulated value. Differentiator 12 is this M
Differentiate the fall of SB output. Therefore, the pulse generation circuit 1 repeatedly generates, as the output of the differentiator 12, a pulse signal (FIG. 3) having a cycle corresponding to the pitch of the formant tone to be generated. This pulse signal is
It is supplied as an initial setting signal to the carrier waveform generating circuit 2 and the window waveform generating circuit 3. Also, the accumulated value output qF _o of the phase accumulator 11 is the compulsory dump waveform memory 4
Is supplied as an address signal.

The carrier waveform generating circuit 2 comprises a phase accumulator 21 and a sine wave memory 22. The phase accumulator 21 accumulates the formant center frequency information value F _f in synchronization with the clock pulse φ, and accumulates the accumulated value qF _f (q = 1,2,3, ...).
Are sequentially output as a read address signal of the sine wave memory 22. The sine wave memory 22 stores the amplitude value sin θ of the sequential sampling points of one cycle of the sine wave at each address, and the address signal (accumulated value q
The sine wave amplitude value stored at the address designated by F _f ) is sequentially read. This allows the sine wave memory 22
From the sinusoidal sequential sampling point amplitude value sin 2πf at the frequency f _f corresponding to the frequency information value F _f according to the clock pulse φ
_f t are output sequentially.

Here, the formant center frequency information value F _f is set to a value corresponding to the formant center frequency f _f which is one of the parameters indicating the tone color of the formant tone independently of the pitch f _{o of the} formant tone to be generated. Has been done. Therefore, the carrier waveform generating circuit 1 is independent of the pitch of the formant tone to be generated, and has a sine wave (see FIG. 3) having a frequency equal to the formant center frequency f _f arbitrarily set corresponding to a desired tone color. ) Will occur.

The window waveform generation circuit 3 includes a phase accumulator 31.
And a sine-square wave memory 32. The phase accumulator 31 accumulates the bandwidth value B _W in synchronization with the clock pulse φ, and the accumulated value q B _W (q = 1,2,3, ...) Is the read address of the sine-square wave memory 32. The signals are sequentially output. Further, as shown in FIG. 3, the phase accumulator 31 stops the accumulation operation when the accumulated value reaches a value corresponding to 2π and continues to output the 2π equivalent value. The sine-squared wave memory 32 stores the amplitude values of sequential sampling points of 1/2 cycle of the sine-squared wave sin ² θ at each address, and is specified by the address signal (accumulated value qB _W ) output from the accumulator 31. The sine square wave amplitude value sin ² kt stored at the designated addresses is sequentially read (FIG. 3).

Waveform memory 4 for forced dump has 1/4 of (1-sin ²ⁿ θ)
Sequential sampling point amplitude values of the cycle are stored in each address, and stored in the address x / 4 specified based on the address signal (accumulated value qF _f = x) output from the accumulator 11 of the pulse generation circuit 1. The forced dump waveform amplitude values that have been read are sequentially read. As a result, the frequency information value is output from the forced dump waveform memory 4 according to the clock pulse φ.
Sequential sampling point amplitude values 1-sin ²ⁿ πf _o t / 2 of the forced dump waveform of frequency f _o corresponding to F _o are sequentially output (third
Figure).

In the multiplier 5, the sine wave amplitude value sin 2πf _f t (FIG. 3) output from the carrier waveform generation circuit 2 and the sine square wave amplitude value si output from the window waveform generation circuit 3
n ² kt (Fig. 3). As a result, the multiplier 5
As the output of, the product of (sin 2πf _f t) and (sin ² kt) (Fig. 3) is obtained.

The multiplier 6 outputs the product (sin 2πf _f
t) · (sin ² kt) is multiplied by the forced dump waveform (1-sin ²ⁿ πf _o t / 2) output from the memory 4. This allows
As the output of the multiplier 6, the first waveform (sin 2πf _f t), the second waveform (sin ² kt) and the third waveform (1-sin ² nπf _o t /
2) and the product (Fig. 3) are obtained. This product indicates the amplitude value of the sequential sampling points of the formant signal.

The multiplier 7 is to add an envelope to the formant signal, and outputs the output of the multiplier 5 (sin 2πf _f
t) · (sin ² kt) · (1−sin ² nπf _o t / 2) is multiplied by the envelope waveform data ENV. As a result, the output of the multiplier 7 is the product of the formant sound amplitude value and the envelope waveform value ENV (sin 2πf _f t) · (sin ² kt)
(1-sin ² nπf _o t / 2) · ENV is obtained. This product indicates the amplitude value of the sequential sampling points 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.

The 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 when 4kπ ≦ x <(4k + 2) π, where k is a positive integer When (4k + 2) π ≦ x <(4k + 4) π In the figure, the parameter n is 1, respectively.
Curve when set to 4,16,64,256 Indicates.

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

Fifth to seventh figure set equal to the window width W and the basic pitch 1 / f _o, formant sound waveforms in the case where the parameter n, respectively 0,16,64 (a) and an amplitude spectrum (b) Indicates. n = 0 means that no forced dump is performed. By comparing these FIG. 5 to FIG.
It can be seen that if the window width W is shorter than the basic pitch 1 / f _{o, the} effect of multiplying the envelope waveform for forced dumping hardly appears.

Eighth to FIG. 10, the base pitches 1 / f _o of the window width W
The waveform (a) and the amplitude spectrum (b) of the formant tone when the parameter n is set to 0 (no forced dump), 16, and 64, respectively, are shown below. By comparing these FIGS. 8 to 10, when the window width W is longer than the basic pitch 1 / f _o , a good formant with a narrow band width can be obtained by multiplying by the waveform for forced dump.

In the above embodiments, the sine squared wave is used as the window waveform, but it is also possible to use a sine 2s squared wave which is a power of this. In this case, the formant band width B _W can be adjusted by changing s. Further, by making s different between the first half and the second half of the window waveform and making the window waveform asymmetric between the first half and the second half, it is possible to adjust the balance between the formant bandwidth B _W and the skirt waveform. Further, by setting s in the latter half to be large, it is possible to reduce the influence of the forced dump.

Further, in the above description, a sine wave is used as the carrier waveform in order to synthesize the formant sound waveform of the unimodal spectrum, but by using a signal having a frequency component corresponding to each peak of the amplitude spectrum as the carrier waveform. , A formant sound waveform having a plurality of peaks in the amplitude spectrum can be synthesized.

Furthermore, in the above description, multiplication is performed using a multiplier, but by treating each data logarithmically, an adder can be used instead of the multiplier. In this case, a logarithm / antilogarithm converter is finally required, but the calculation speed can be increased.

[Effects of the Invention] As described above, in the conventional case, after adding each phoneme by using two or more operators, the adder adds the phonemes so that the width W of the window waveform is long (formant band width is However, according to the present invention, each time a new phoneme starts, if the previous phoneme continues, it will be forcedly dumped. It is possible to synthesize two formant sounds. Therefore, although a waveform generator for forced dump is required, the number of operators can be reduced,
As a result, the hardware configuration can be simplified. Alternatively, a larger number of formant sounds can be simultaneously generated with the same hardware configuration. in this case,
The longer the window width W with respect to the basic pitch 1 / f _o ,
The effect of forced dumping appears strongly, eliminating the need for complicated control logic such as generating phonemes in a cycle of f _o / n (n is the number of operators) and accumulating them, and also possible formants Greater advantages such as an increase in the number of sounds can be obtained.

Further, even when the forced dump is always performed, when the window width W of the phoneme is short, the influence of the forced dump is small, and it is possible to sufficiently obtain the same performance as the conventional one. Is.

In addition, the conventional method cannot handle phonemes with a period longer than f _o / n. Such phonemes are
This often occurs when the formant width is narrowed in the high tone range, but according to the present invention, the formant tone can be synthesized without being subject to such a limitation.

[Brief description of drawings]

FIG. 1 is a waveform diagram for explaining the basic operation of the invention, FIG. 2 is a block diagram of a formant sound synthesizer according to the invention, and FIG. 3 is an operation timing of each part of the apparatus of FIG. 4 and FIG. 4 are envelope waveform diagrams for forced dump generated in the apparatus of FIG. 2, and FIGS. 5 to 10 are formant sounds synthesized when various parameters of the apparatus of FIG. 3 are set variously. FIG. 11A is a waveform diagram, FIG. 11B is an amplitude spectrum diagram, and FIGS. 11A and 11B are waveform diagrams and amplitude spectrum diagrams showing an example of natural formant sounds. , (A) and (b) are schematic spectrum diagrams of natural formant sounds, FIGS. 13 (a) and (b) are diagrams showing carrier waveforms and window waveforms used in a conventional formant synthesizer, and FIGS. 14 (a) and (b) are main diagrams. The inventor explained the operation of the method that he tried first. Which is the window waveform diagram. 1: Pulse generation circuit 2: Carrier waveform generation circuit 3: Modulation (window) waveform generation circuit 4: Forced dump waveform memory, 5, 6: Multiplier 33: Sine square wave memory

Claims (2)

(57) [Claims] 1. A predetermined first set arbitrarily corresponding to a desired tone color independently of the pitch of a formant tone to be synthesized.
A first waveform generating means for generating a first waveform having a repetition period of, and a sine wave squared or exponentiated waveform thereof.
Second waveform generating means for generating a second waveform having a second cycle longer than the repetition cycle of, and pulse signal generating means for repeatedly generating a pulse signal having a cycle corresponding to the pitch of the formant tone, The waveform generating operation in the first and second waveform generating means is repeatedly initialized by a pulse signal, and the first and second waveforms are generated at a cycle corresponding to the pitch of the formant tone.
Control means for repeatedly generating the waveform of, and at least when the cycle of the pulse signal is shorter than the second cycle, the third cycle of the pulse signal starts from 1 in synchronization with the pulse signal and reaches 0 smoothly. And a multiplication means for multiplying the waveforms generated by the first to third waveform generation means, and a formant sound waveform is generated as an output of the multiplication means. Formant sound synthesizer characterized by the above. 2. The third waveform, when x is a phase and k is a positive integer and 4kπ ≦ x <(4k + 2) π When (4k + 2) π ≦ x <(4k + 4) π The formant sound synthesizer according to claim 1, which has a waveform represented by