JPH0632000B2 - Speech tone synthesizer - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a voice (musical sound) synthesizer, and more particularly to a device that synthesizes sounds such as voices, singing voices, and musical tones at a precise pitch.

(Prior Art and its Problems) Conventionally, as a device for synthesizing voices and musical sounds, which is realized by a sampled value system, it is composed of a means for generating a sound source waveform signal and a synthesizing filter with variable frequency characteristics. In the case of, a periodic sound source waveform signal such as an impulse train is generated, in the case of unvoiced sound, a random sound source waveform signal is generated, and the synthesis filter is controlled based on the information indicating the spectral envelope to generate the sound source. It is known to wave a waveform signal to synthesize a voice or musical sound.

For example, JD. Markel and J. H. Co-authored by AHGray, `` Linear Prediction of Speech (Linear Prediction o
The above example is shown in Chapter 1 of "f Speech)".

According to the above-mentioned conventional example, since the voice is synthesized from the information indicating the feature of the sound source and the information indicating the spectrum envelope, it is possible to synthesize the voice with a small amount of information, and it is also suitable for a device that synthesizes the voice by the rule. There are features.

However, in the above-mentioned conventional example, the pitch of the periodic sound source waveform signal can be generated only at a cycle which is an integral multiple of the sampling cycle, and therefore when synthesizing a singing voice or a musical sound, a synthesized sound having an accurate pitch cannot be obtained. There was a flaw. Furthermore,
In order to improve the accuracy of the pitch, the sampling cycle must be shortened, that is, the sampling rate must be increased, which requires a lot of calculation. These drawbacks become a problem especially when synthesizing sounds with a high pitch.

(Object of the Invention) It is an object of the present invention to provide a voice (tone) synthesizer capable of synthesizing a voice or a tone having a more accurate pitch than a conventional one while having a simple circuit configuration.

(Structure of the Invention) A speech sound synthesizing apparatus according to the present invention comprises means for generating a periodic sound source waveform signal, and a synthesis filter having variable frequency characteristics controlled based on spectrum envelope information. In a device for synthesizing a voice or a musical sound by oscillating with the synthesis filter with a low pitch between the output side of the synthesis filter and one fixed contact of the changeover switch Of the fundamental frequency of the sound and its harmonic components, a comb filter with a variable pass band that passes only the desired pitch frequency and its harmonic components is connected, and the output side of the synthesis filter and the other of the changeover switch are connected. The fixed contact is directly connected, and the output is taken out from the movable contact side of the changeover switch.

(Principle of the Invention) The present invention generates a sound source waveform signal having a long period, and controls the synthesis filter based on the spectrum envelope information to wave the sound source waveform signal to generate a synthesized sound having a long period, that is, a low pitch. To generate. Of the fundamental frequency and its harmonic components of the low pitch synthetic sound, the low pitch synthetic sound is controlled while controlling the pass band of the comb filter so that only the desired pitch frequency and its harmonic components are passed. To generate a synthesized sound with a desired pitch.

By using a periodic sound source waveform signal having a long period, the pitch accuracy is relatively improved. Then, if the comb frequency filter suppresses the unnecessary frequency component of the synthesized sound that is once generated at a low pitch, even if the unnecessary frequency component cannot be completely removed, a synthesized sound with a high pitch can be obtained in terms of hearing.

(Embodiment) Next, an embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram of an embodiment of the present invention. In the figure,
2 is an impulse train generation circuit for generating a periodic sound source signal, 3 is a random noise generation circuit, 5 is a synthesis filter, 6
Comb filters 4 and 7 are a first switch and a second switch, respectively, and 1 is a control circuit.

Based on the information input from the signal line 8, in the case of a periodic sound such as a voiced sound, the control circuit 1 transmits the period information and the amplitude information of the sound source to the impulse train generation circuit 2 via the signal line 10. The selection information of the switch 4 of No. 1 is sent to the switch 4 via the signal line 12, the spectrum envelope information is sent to the synthesis filter 5 via the signal line 13, and the pass band control information is sent to the signal line 14.
The selection information of the second switch is sent to the comb filter 6 through the switch 7 and the switch 7 to the switch 7 through the signal line 15.

The impulse train generated by the impulse train generating circuit 2 passes through the signal line 16 and is selected by the first switch 4, and is sent to the synthesis filter 5 through the signal line 18 to be waved.
The output of the synthesizing filter 5 is passed through a signal line 19 and unnecessary frequency components are suppressed in the comb filter 6 to generate a synthesized sound, which is selected by the second switch 7 through the signal line 20 and output from the signal line 9. To be done.

However, at this time, when a desired pitch accuracy can be obtained without using the comb filter 6 because the desired pitch period is long, the control circuit 1 instructs the switch 17 to
Information for instructing to select the side of the signal line 21 is sent, and the output waveform signal of the synthesis filter 5 is output as it is through the signal line 21, the switch 7, and the signal line 9.

On the other hand, in the case of non-periodic voice such as unvoiced sound, the amplitude information of the sound source is sent from the control circuit 1 to the random noise generation circuit 3 via the signal line 11, and the selection information of the first switch 4 is sent to the signal line 12. The spectral envelope information is sent to the switch 4 via the signal line 13 to the synthesis filter 5 and the selection information of the second switch 7 is sent to the switch 7 via the signal line 15.

The random sound source waveform signal generated by the random noise generator 3 passes through the signal line 17, is selected by the first switch 4, is sent through the signal line 18 to the synthesis filter 5, and is waved. The output of the synthesis filter 5 passes through the signal line 21, is selected by the switch 7, and is output from the signal line 9.

(Effects of the Invention) As described above, the present invention generates a sound source waveform signal having a long cycle, and oscillates the sound source waveform signal while controlling the synthesis filter based on the spectrum envelope information to obtain a long cycle, that is, a pitch. A synthetic filter with a low pitch is generated, and the pass band of the comb filter is set so that only the frequency of the desired pitch and the harmonic component of the fundamental frequency and its harmonic components of this low pitch synthetic sound are passed. Since the synthesized sound having a low pitch is generated by controlling the synthesized sound having a low pitch, the pitch accuracy can be relatively improved by using a periodic sound source waveform signal having a long period. .
Then, by suppressing the unnecessary frequency component of the synthesized sound once generated at a low pitch with the comb filter, it is possible to obtain a synthesized sound with a high pitch even if the unnecessary frequency component cannot be completely removed. effective.

Further, since the present invention is the same as the conventional structure except for the comb filter, the present invention can be easily realized by externally attaching the comb filter to the voice synthesizer made by the conventional structure. This configuration is effective for improving the pitch accuracy in the case of a speech synthesizer or the like which has already been integrated into an LSI.

[Brief description of drawings]

FIG. 1 is a block diagram of an embodiment of the present invention. 1 ... Control circuit, 2 ... Impulse train generation circuit, 3 ...
Random noise generation circuit, 4, 7 ... Switch, 5 ... Synthesis filter, 6 ... Comb filter.

Claims (1)

[Claims] 1. A means for generating a periodic sound source waveform signal,
An apparatus for synthesizing a voice or a musical tone by synthesizing a sound source waveform signal by the synthesis filter, comprising: a synthesis filter having variable frequency characteristics controlled based on spectrum envelope information; Between one of the fixed contacts and the fundamental frequency of the synthesized sound with a low pitch of the sound source waveform signal waved by the synthesis filter and its harmonic components, pass only the desired pitch frequency and its overtone components. A comb-shaped filter having a variable band is connected, the output side of the synthesis filter and the other fixed contact of the changeover switch are directly connected, and the output is taken out from the movable contact side of the changeover switch. A voice music synthesizer.