JPS58215700A - Voice synthesizer - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a speech synthesis device, and particularly to a synthesis device that requires sound source amplitude data as a synthesis parameter.

In general, the sound source amplitude parameter for each predetermined section used to set the initial value for synthesis in a parameter synthesis type speech synthesizer is, for example, a natural speech waveform a as shown in the first factor, and the following formula: The average amplitude value that can be determined by is used.

Here, A is the average amplitude value, n is the number of samples in a certain interval e
”(ri indicates the amplitude value at time i.

When synthesizing speech using such an average amplitude value, for example, in conventional formant synthesis, the characteristics 2 of the synthesizer, for example,
Because the average amplitude data of natural speech (original speech) derived from the above formula was used as is, without taking into account the suitability of other speech parameters or errors during their quantization, the reproduced synthesized There were drawbacks such as the sound exhibiting an abnormal amplitude value, deterioration of the sound quality, and difficulty in controlling the sound level.

An object of the present invention is to provide a synthesizer '1 that does not cause deterioration of sound quality or level fluctuation.

The speech synthesis device of the present invention has means for correcting the amplitude average value extracted from the original speech so as to be suitable for the synthesis device, and the speech is synthesized using the corrected amplitude average value. Features.

According to the present invention, there is provided a means for synthesizing speech using a predetermined value as a sound source amplitude value, a means for extracting an average amplitude value of the synthesized speech, and a speech synthesized with the average amplitude value of natural speech. and means for correcting all of the predetermined values according to the output of the compared results, and the corrected value is thereby calculated as the sound source amplitude of the speech synthesizer. It is characterized by being used as a value.

According to the present invention, when synthesizing speech, natural speech (original speech)
Rather than using the average amplitude value extracted from the source as the source amplitude value as it is, we correct it to suit the synthesizer before using it. The above correction compensates for deterioration of sound quality and fluctuations in level caused by other audio parameters (for example, voicing parameters and pitch parameters), errors in quantizing the parameters, or characteristic errors of the synthesizer itself used. can. Therefore, no abnormal amplitude or deterioration of sound quality is observed in the synthesized speech. Furthermore, if a synthesizer uses a similar parameter synthesis method, even if there is a characteristic error that causes the amplitude average value of the device itself to fluctuate, it will be possible to effectively compensate for all of these errors and synthesize speech. can.

Hereinafter, one embodiment of the present invention will be described in detail with reference to FIG. 2.

FIG. 2 is a functional diagram of an embodiment in which the present invention is applied to a formant synthesis device. In the figure, 1 is a speech analysis device, 2 and 4 are average amplitude value extraction devices, 3 and 6 are speech synthesis devices, and 5 is a comparator. In addition, 7 is a natural voice input, and 8 is a parameter other than the sound source amplitude analyzed from the natural voice (voice parameter, voice presence/absence parameter, and pitch/voice parameter).
Parameter) 9 is the average amplitude value obtained from the original voice, 10
is a sound source amplitude parameter with a predetermined value (
(Note that all the average field of view values obtained from natural voices may be used as they are.) 11 is a synthesized voice.

12 is the average amplitude value of the synthesized speech 11, 13 is 9.10°1
The sound source amplitude parameter 14 obtained by comparing the average amplitude values of 2 is the target synthesized speech.

However, the average amplitude extraction devices 12 and 4. Or speech synthesizer #:
, 3 and 6 may share exactly the same thing, but
They are shown in separate block diagrams to make the explanation easier to understand.

The operation based on this configuration will be explained below. First, natural speech 7 is input to a speech analysis device R1 equipped with an average amplitude value extraction device 12, and an average amplitude value 9 and other parameters (speech parameters, etc.) 8 are obtained as output.

The various parameters 8.9 are not directly used for final speech synthesis, but the sound source amplitude parameters are adjusted in advance. In the adjustment, the sound source amplitude parameter 10 of each building parameter 8.9 is temporarily input to the same speech synthesizer 3 as the speech synthesizer 6 using a certain other predetermined value to synthesize speech. View. The average amplitude value 12 of the output synthesized speech 11 thus obtained is obtained by the average amplitude value extraction device 4, and is used as the average amplitude value 9°10.12 to be used in the comparator 5.
The optimum sound source amplitude parameter 13 for this speech synthesizer 6 is determined by comparison. As a result, by using this optimum sound source amplitude parameter 13 and other various parameters 8, the synthesized speech 14 having an amplitude value faithful to the amplitude value of the natural speech 7 can be obtained by the speech synthesizer 6.

Next, this comparator 5 will be explained. We want the average amplitude value of the synthesized speech 14 obtained by the speech synthesizer 6 to be the same as the average amplitude value 9 of the original speech 7, but if we assume that the parameters other than the sound source amplitude parameter are the same, now the speech synthesis When an average amplitude parameter having a predetermined value is input to the device 3, an average amplitude value 12 of the synthesized synthesized speech-11 can be obtained. Therefore, in order to obtain the synthesized speech 14 that is the same as the average amplitude value 9 of the natural speech 7, the following sound source amplitude parameter 13 may be outputted from the comparator 5 as an input to the speech synthesizer 6.

That is, considering all the gains of this speech synthesizer, let the sound source amplitude parameter 10 having a predetermined value be E, and Ei
The average amplitude value of the voice 11 synthesized using
The average amplitude value of the natural voice is 'ID', and the value that the comparator 5 should output (CD/ When the exponent value is 13+E/, the following formula holds true.

D/E=D'/E' Therefore E' = D' x E/D That is, the comparator 5 calculates D' and E based on this formula.

It may be configured to receive input of D and output E'.

As explained above, the present invention takes into account the compatibility of the sound source amplitude parameter extracted by the speech analysis device with other parameters and the characteristics of the speech synthesizer, and corrects the sound source amplitude parameter to the optimal one. This has the great effect of faithfully reproducing the amplitude value of the audio.

[Brief explanation of drawings]

FIG. 1 is a waveform diagram of audio in a certain period. a: Number of samples of audio waveform n knee section a(0), a(1)...: Time 0.1...
. . . Amplitude values FIG. 2 is a functional diagram showing an embodiment of the present invention. 1... Speech analysis device, 2.4... Average amplitude value extraction device, 3.6... Speech synthesis device, 5.
...Comparator, 7...Natural voice input terminal,
8... Parameters other than sound source amplitude, 9...
...Average amplitude value obtained from natural speech, 10...
・Sound source amplitude parameter with a predetermined value (an average amplitude value obtained from natural speech may be used), 1
1... Synthesized speech, 12... Pa... Average amplitude value of synthesized speech of 11, 13... Sound source amplitude obtained by comparing the average amplitude value of 9° 10.12. Parameter, 14
...Synthesized audio output terminal.

Claims (1)

[Claims] means for synthesizing speech using speech parameters extracted from natural speech and predetermined sound source amplitude value parameters; means for extracting an average amplitude value of the synthesized speech; and an average amplitude value of the natural speech. and a means for comparing the average amplitude value of the synthesized voice and correcting the predetermined value accordingly, and using the corrected value as a sound source amplitude value and using this and the voice parameter to generate the target voice. 1. A speech synthesis device comprising: means for synthesizing.