JPH0646356B2 - Sound source signal generation method - Google Patents

Description

The present invention relates to a method of generating a sound source signal for high-quality sound synthesis used when synthesizing sounds such as voices and musical tones using filters.

When synthesizing sounds such as voices and musical instrument sounds using a filter, a periodic driving sound source signal is used.

Thus, conventionally, as such a sound source signal, an impulse train, a pulse train obtained by passing the impulse train through an all-pass filter, a pulse train having a repeated Baker sequence, etc. are used.

In addition, a periodic signal is also used, in which a signal obtained by passing an audio through an inverse filter having an inverse characteristic of its spectrum envelope is cut out and repeated from a waveform of a typical specific section of a so-called residual.

By the way, according to the law about the tone color of the musical sound of Helmholtz, it is considered that it is enough to use a simple impulse train as the sound source signal. It has been suggested that this Helmholtz law is not necessarily exact, but it is not known what is the more appropriate signal other than the impulse train. For this reason, the above-described impulse train has been the most basic sound source signal. However, the use of such impulse trains has a drawback in that the quality of synthesized sounds cannot be obtained. On the other hand, the one using the typical residual is superior to the one using the impulse train in terms of sound quality, but it is difficult to control the fundamental frequency, and the quality of the synthesized sound is not always good depending on the kind of sound. was there.

The present invention has been made to eliminate the above drawbacks, and an object of the present invention is to provide a sound source signal generation method capable of obtaining a sound source signal for high quality sound synthesis.

An embodiment of the present invention will be described below.

First, as shown in FIG. 1 (a), the amplitude is A and the cycle is NΔt (Δt
Is a unit time)), the amplitude spectrum of such an impulse train has a fundamental frequency f ₀ (= 1
A group of line spectra having a uniform amplitude of Af ₀ and having a cycle of / (NΔt)). This is called a basic impulse train.

Considering an impulse train having an amplitude of B and a period of NΔt / 2 shown in FIG. 1 (b), the amplitude spectrum of such an impulse train has a size with a period of 2f ₀ starting from frequency 0. Is a line spectrum group having a uniform amplitude of 2Bf ₀ . In this case, the impulse train of FIG. 1 (b) is the one without the odd harmonic components, considering the amplitude spectrum of the impulse train of FIG. 1 (a). Call.

Next, the amplitude shown in FIG. 1 (c) is alternately ± B and the cycle is NΔ.
Considering a t / 2 impulse train, the amplitude spectrum of such an impulse train is a group of line spectra with a uniform amplitude of 2Bf ₀ with a period of 2f ₀ starting from the frequency f ₀ . In this case, the impulse train in Fig. 1 (c) is the same as that in Fig. 1 (a).
Considering the amplitude spectrum of the impulse train as a reference, the even harmonic components are removed, and hence it is called the odd harmonic fundamental impulse train here.

Then, of the pulse trains having such a relationship, FIG. 1 (b)
Items (c) and (c) are superimposed in an arbitrary time relationship. This is shown in FIG.

In this case, it is the same as the even harmonic fundamental impulse train in Fig. 1 (b).
Since the frequency components do not overlap with each other in the odd-harmonic fundamental impulse train in (c), and the components have the same magnitude, the pulse train in FIG. 2 has a first amplitude spectrum if the amplitude B is A / 2. It is completely equal to the basic impulse train in Fig. (A). On the other hand, however, the pulse train of FIG. 2 is significantly different from the basic impulse train of FIG. 1 (a) in the phase and the structure of the time transition of the amplitude spectrum in a time section shorter than the period of the pulse train (hereinafter referred to as the time structure). That is, in the pulse train shown in FIG. 2, the short-time amplitude spectrum alternates between the low-frequency emphasis section and the high-frequency emphasis section in one period section of the waveform.

Therefore, according to the sound source signal given by such a pulse train, the time structure of the time spectrum of the short-time spectrum of the synthesized sound becomes close to that of natural sound, and as a result, compared with various conventionally used pulse train sound sources, the synthesized sound The quality of can be dramatically improved.

If N is not divisible by 2 in the generation of the pulse train shown in FIG. 2, a close integer may be selected as the value of N / 2.

Next, the period NΔt of the pulse train shown in FIG. 2 is changed to NΔt / 2.
Considering the pulse train shown in FIG. 3 in such a case, the amplitude spectrum of the pulse train at this time is a line spectrum group of uniform amplitude with a frequency of 2f ₀ as a starting point and a period of 2f ₀ and a size of 4Bf _0. . That is, the pulse train of FIG. 3 is an even harmonic pulse train having no odd harmonic components.

In this case, however, such a pulse train and an odd-numbered impulse train having an amplitude of 2B are superposed with an appropriate time relationship. Then, the waveforms at this time are shown in FIGS. 4 (a), (b), and (c), and the amplitude spectrum is the same as that of the basic impulse train when B is A / 4.

Therefore, in the pulse trains of FIGS. 4 (a) (b) (c) as well, the high-frequency emphasis section and the low-frequency emphasis section of the short-time spectrum appear in one period section, and as a sound source for high-quality sound synthesis. It will be available.

The pulse train shown in FIG. 4 is the same as the pulse train shown in FIG.
The odd-harmonic basic impulse train was superposed on the even-harmonic pulse train obtained by halving, but in the same way, based on the pulse train in Fig. 4, a new amplitude spectrum was added to the basic impulse train. If a pulse train equivalent to is generated, the above procedure can be repeated in principle, and a variety of sound source signals for high quality sound synthesis can be obtained.

Therefore, by doing so, it is possible to provide a sound source signal generation method capable of obtaining a sound source signal for high-quality sound synthesis.

The present invention is not limited to the above-described embodiments, and can be implemented by appropriately modifying it without departing from the scope of the invention. For example, by making the magnitude of the even harmonic component smaller than the magnitude of the odd harmonic component, it is possible to generate a sound source signal for synthesizing the sound of a closed wind instrument such as a clarinet.

[Brief description of drawings]

FIG. 1 (a) (b) (c), FIG. 2, FIG. 3 and FIG. 4 (a) (b)
(c) is a waveform diagram for explaining one embodiment of the present invention.

Claims (1)

[Claims] 1. A pulse train of a sound source signal is obtained by superimposing a pulse train containing almost no odd harmonic components of a fundamental frequency and a pulse train containing almost no even harmonic components at an arbitrary time relationship. Characteristic sound source signal generation method.