JPS5848920B2 - Speech synthesizer sound source creation device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a sound source creation device for a speech synthesizer that performs gradual changes in short-term amplitude and continuous voiced/unvoiced switching operations on the sound source of the speech synthesizer.

A conventional voice synthesis device is constructed as shown in FIG.
Also, the sound source creation circuit is structured as shown by the dotted line in FIG.

In FIG. 1, 1 is an audio information manual terminal, 2 is an audio output terminal, 3 is an audio information separation circuit, 4 is an output terminal for spectrum information, 5 is an output terminal for voiced/unvoiced switching information, 6 is a pitch information output terminal, 7 is an output terminal for amplitude information, and the signal inputted as a serial signal from input terminal 1 is separated into output information from terminals 4, 5, 6, and 7 of audio information separation circuit 3, and these outputs have a frame period. (about 10 msec~
It is updated every 25 msec).

8 is a counter, and this count value is 1 from the pitch period value.
It is decremented by 1 every 25 microseconds, and when it reaches 0, it is taken as the value of the pitch period again. Only when the count value is O, the counter output becomes 1, and in other cases, the output is O.

9 is a multiplier, and by multiplying the output of the counter 8 and the amplitude value of the terminal T by the multiplier 9, a pulse of the value specified at the terminal 7 is obtained for each pitch period.

That is, the output of the multiplier 9 provides a sound source for voiced sound.

10 is a random pulse generation circuit that randomly generates 1 and -1 every 125 μs, 11 is a multiplier, and the output of this multiplier 11 generates pulses with irregular polarity depending on the amplitude value specified at terminal 7. obtain.

That is, the output of the multiplier 11 provides a sound source for unvoiced sound.

Reference numeral 12 denotes a changeover switch circuit which selects the information obtained by the multipliers 9 and 11 based on the voiced/unvoiced switching information designated by the terminal 5.

The sound source information obtained by this changeover switch circuit 12 is transmitted to the terminal 4.
A digital audio output is obtained by the synthesis digital filter circuit 13, and a synthesized speech is obtained from the speaker 15 via the digital-to-analog conversion circuit 14.

As shown in Figure 2, the operation of this conventional example results in amplitude discontinuity at the boundaries between frames, a smooth transition from unvoiced sound to voiced sound, and a synthesized sound that differs from actual speech. It had the disadvantage that it could be obtained.

Furthermore, as shown in Figure 3, it has already been proposed that spectral information can be used by interpolation to obtain a combined sound, but in this case, in order to obtain a high-quality synthesized sound, it is necessary to It was necessary to use it as information by shifting it by 1/2 frame.

In order to solve the above-mentioned drawbacks of the conventional example, the present invention performs interpolation on the amplitude of the pulse train and random noise to prevent sudden changes in the amplitude, and at the same time, at the time of switching between voiced and unvoiced, the pulse train and random noise are interpolated. The present invention provides a sound source creation device for a speech synthesizer that enables smooth voiced/unvoiced switching operations and improves the quality of synthesized speech by providing mixed sections.

Hereinafter, embodiments will be described in detail with reference to the drawings.

FIG. 4 shows one embodiment of the present invention, in which the same reference numerals as in FIG. , and outputs O-level amplitude information to terminal 18 . During silence, O-level amplitude information is output to terminal 17 , and amplitude information from terminal 7 is output to terminal 18 .

Here, the information obtained at the terminals 17 and 18 is the amplitude information of the voice component and the amplitude information of the unvoiced component, respectively.

In addition, the spectrum information of terminal 4, the pitch period of terminal 5, the amplitude information of voiced components, and the amplitude information of unvoiced components from terminals 17 and 18 are interpolated by interpolation circuits 19, 20, 21, and 22, respectively, and change smoothly. Used for information.

Furthermore, these interpolation circuits 19, 20, 21, and 22 are similar circuits, and can have a common circuit configuration.
It does not significantly increase the circuit scale.

Also, the amplitude information of the voiced component obtained by multiplying the outputs of the counter 8 and the interpolation circuit 27 by the multiplier 9 and the unvoiced component obtained by multiplying the output of the random noise generator 10 and the output of the interpolation circuit 22 by the multiplier 11 are added together. The signals are synthesized in a circuit 23 and sound source information is obtained.

This sound source information is used in the synthesis digital filter circuit 13 together with the spectrum information from the terminal 4 to obtain a digital audio output, and a synthesized audio is obtained from the speaker 15 via the digital-to-analog conversion circuit 14.

Note that a portion 24 indicated by a dotted line is a sound source production circuit.

In order to explain the operation of the above embodiment, FIG. 5 shows information changes in each part of FIG. 4. In this way, smoothly changing amplitude information of voiced and unvoiced components can be obtained. .

This amplitude information is 1
Although there is a statistical deviation by one /2 frame, by interpolating the spectrum information and pitch information as in this example, the phase becomes the same as that of the amplitude information, and a system in which the phases of each information match is constructed. .

As explained above, according to the present invention, it is possible to realize continuous voiced/unvoiced switching operation and smooth change in amplitude without causing a large increase in circuit scale, and it can be used as a sound source for a speech synthesizer. In this case, it is possible to improve the quality of the synthesized speech, and by also interpolating spectrum information and pitch information, there is an advantage that a system in which the phases of each information are matched can be constructed.

[Brief explanation of drawings]

Figure 1 is a configuration diagram of a conventional sound source creation device;
The figure shows an example of how each information changes over time in the conventional method, Figure 3 shows an example of interpolation of spectrum information, and Figure 4 is a block diagram of one embodiment of the present invention. , FIG. 5 is a diagram showing temporal changes in each piece of information in FIG. 4. 1...Audio information input terminal, 2...Audio output terminal, 3...Audio information separation circuit, 4...
... Spectral information output terminal, 5 ... Voiced/unvoiced switching information output terminal, 6 ... Pitch information output terminal, 7 ... Amplitude information output terminal, 8 ...
... Counter, 9, 11 ... Multiplier, 10...
... Random pulse generator, 13 ... Synthesis digital filter circuit, 14 ... Digital-to-analog conversion circuit, 15 ... Speaker, 16
...branch circuit, 17,18...terminal,
19, 20, 21. 22... Interpolation circuit, 23.
...Addition circuit.

Claims (1)

[Claims] 1 Voiced/unvoiced judgment is performed for each analysis frame, and a pulse train is generated for voiced sounds, and random noise is generated for unvoiced sounds.The amplitudes of voiced sounds and unvoiced sounds are interpolated for each sound source signal. 1. A sound source creation device for a speech synthesizer, characterized in that a mixed section of a pulse train and random noise is set up.