JPS62150398A - Voice rule 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 (Technical Field) The present invention relates to a speech rule synthesis device that can synthesize arbitrary words.

(Prior Art) Generally, speech rule synthesis for uttering an arbitrary word is performed in the following steps.

■ Input the sentence you want to utter as a character code string (Kana, Romaji, etc. code string), and then create a sound unit smaller than a word, such as a syllable (generally, if you exclude single vowels such as vowels in consonants, 2 0 ■ Corresponding to each column of the synthesis basic unit, name of the synthesis basic unit that is stored in advance (hereinafter referred to as synthesis basic unit name) string Audio parameters (spectrum/
parameters, pitch parameters, sound source amplitude parameters, etc.) 0 ■ The audio parameters are combined based on predetermined rules to synthesize audio as a sentence.

In this way, any womb can be synthesized.

However, when combining audio parameters between synthesis basic units in step This is important for smoothly combining audio parameters at basic unit boundaries.

This conventional method will be explained in more detail with reference to FIGS. 1 and 2. FIG. 1 is a block diagram showing an example of a speech rule synthesis device to which a conventional method is applied. In the same figure,
10 is a character code string input circuit for inputting character code strings using keys, etc.; 20 is a character code string-composite basic unit name string conversion circuit for converting the character code string into a composite basic unit name string; 3
0a is an audio parameter read control circuit that outputs the storage address of the audio parameter corresponding to the synthesis basic unit name string, 40 is an audio parameter memory in which audio parameters are written corresponding to the synthesis basic unit name, and 51 is an audio Either send the parameters directly to the speech synthesis circuit 60 or send them to the interpolation circuit 52a.
Audio parameter control switch 52a for selecting whether to send to
53a is an interpolation circuit that interpolates audio parameters at the synthesis basic unit boundary; 53a is a combination control circuit that controls the overall braking of the audio parameter combination circuit 50a; and 50a is 51, 52a.
, 53a, an audio parameter combination circuit 60
70 is a voice synthesis circuit that synthesizes voice using voice parameters, and 70 is a voice output circuit consisting of a low bass filter, an amplifier, and a speaker.

FIG. 2 is a block diagram illustrating the input section of the interpolation circuit 52a of FIG. 1 in more detail. In the same figure, 5
5 is an interpolator that performs interpolation between buffers 56 and 57; 56 is a buffer that inputs the audio parameters of the final part in the preceding basic synthesis unit; and 57 is the audio parameter of the first part in the subsequent basic synthesis unit. Enter 'Rose'
7% 58 inputs audio parameters to buffer 56?

This is a switch for selecting whether to input data to the buffer 57.

Next, the operation based on this configuration will be explained.

First, input the sentence to be uttered as a character code string into the character code string input circuit lO.Next, input the sentence as a character code string into the character code string
The composite basic unit name string conversion circuit 20 converts the input character code string into a composite basic unit name string. The audio parameter read control circuit 30a generates an address on the audio parameter memory 40 storing the audio parameter corresponding to the basic synthesis unit name. Then, using the address, the voice parameters are read from the voice parameter self-memory 40, and the voice is synthesized by the voice synthesis circuit 60,
It is output from the speech synthesis circuit 70.

On the other hand, when combining audio parameters at the boundary of each basic synthesis unit, the audio parameter readout control circuit 30a stops reading the audio parameters, and the combination control circuit 53
A controls the switch 51, the interpolation circuit 52a, and the switch 58, inputs the audio parameters of the last part in the preceding synthesis basic unit to the buffer 56 through the switch 58, and inputs the audio parameters of the first part in the following synthesis basic unit. Parameters are input into buffer 57 through switch 58.

Interpolator 55 performs interpolation between buffer 56 and buffer 57. Then, the speech synthesis circuit 60 synthesizes speech using the smoothly combined speech parameters, and the speech output circuit 70 outputs the speech.

In this way, in general, speech with excellent naturalness can be synthesized by smoothly combining vowels and vowels, vowels and fricatives, etc.

However, if the amplitude of the first phoneme in the subsequent synthesis basic unit is characterized by a rapid change (as an example of a steep rise, the waveform of 1T@ in a sentence is shown in Figure 3),
As mentioned above, interpolation of the parameter related to the sound source amplitude in the same manner as the other parameters has the disadvantage that the clarity of the phoneme deteriorates.

(Object of the Invention) An object of the present invention is to provide a speech rule synthesis device that can reproduce phonemes characterized by rapid changes in amplitude more faithfully than ever before.

(Structure of the Invention) A means for inputting an arbitrary sentence as a character code string, a means for converting the input character code string into a synthetic basic unit name string, and a voice parameter (
Spectral parameters.

a sound parameter memory that stores sound source amplitude parameters, pitch meeting parameters, etc.) in advance, means for reading out sound parameters from the synthesis basic unit name sequence, and combining means for interpolating and combining the sound parameters of the synthesis basic units. and means for synthesizing speech using the combined speech parameters,
The combining means includes a storage means 1 for temporarily storing the sound source amplitude parameter of the last part of the preceding basic synthesis unit, a temporary storage means 2 for the sound source amplitude parameter of the first part of the subsequent basic synthesis unit, and a predetermined sound source. constant value generation means for generating an amplitude value; selection means for selecting and outputting the output of the storage means or the output of the constant value generation means according to the name of the subsequent synthesis basic unit; and the output of the selection means l; This configuration includes interpolation means for interpolating the output of the storage means l.

(Operations and Effects of the Invention) Interpolating between each basic unit of synthesis allows voice parameters to be smoothly combined and a synthesized sound with excellent naturalness can be obtained, but it is also possible to interpolate between the input character code strings. , phonemes whose amplitude changes rapidly (e.g. I Tl , @p
1, @K 1, etc.), by once reducing the amplitude of the preceding phoneme to O, the amplitude of the following phoneme can be raised steeply, and the natural speech waveform can be faithfully reproduced. It has the effect of being able to be regenerated.

(Example) Examples of the present invention will be described below along with FIGS. 4 and 5.
Let me explain in detail.

FIGS. 4 and 5 are block diagrams showing an embodiment of a speech synthesis device suitable for the present invention. In the same figure, the parts given the same numbers as in FIGS. 1 and 2 have the same functions, so their explanation will be omitted here. In addition, in the same figure, 54 is a sound source amplitude parameter generation circuit that generates a sound source amplitude parameter whose sound source amplitude is O, and 30b and 5
3b is the audio parameter readout control circuit 30a in FIG.
A circuit for controlling the sound source amplitude parameter generation circuit 54 is added to the combination control circuit 53a. In the interpolation circuit 52b of FIG. 5, 59 is a switch for selecting whether to input the contents from the buffer 57 to the interpolator 55 or the value generated from the sound source amplitude parameter generation circuit 54. (Since the circuits were changed as described above, the interpolation circuit 52a was changed to 52b, and the audio parameter combination circuit 50a was changed to 50b).

Next, the operation based on this configuration will be explained.

First, an arbitrary sentence, for example l'-AISATUJ, is input to the character code string input circuit IO. Next, the character code string-synthesis basic unit name string conversion circuit 20 receives the input l”Al
5ATUJ to rAJ, r month, ``8AJ.

It is converted into a synthetic basic unit name string for each syllable, such as "TU". In the audio parameter readout control circuit 30b,
Its rAJ, rIJ, r8AJ, rTUJ
An address on the audio parameter memory 40 storing the audio parameters corresponding to the respective synthesis basic unit names is generated. Then, using that address, the voice parameters are read from the voice parameter memory 40, changed to voice by the voice synthesis circuit 60, and the voice parameters are changed to voice by the voice output circuit 70.
Output from.

On the other hand, at the boundary of each basic unit of synthesis, for example, rAJ and rIJ
When combining audio parameters at the boundary of The parameters and the audio parameters of the first part of the following rIJ are inputted to the interpolation circuit 52b through the switch 51. In the interpolation circuit 52b, the preceding "
The audio parameters of the last part of "A" are inputted into the buffer 56 through the switch 58, and the audio parameters of the first part of the following "work" are inputted into the buffer 57 through the switch 58.
An interpolator 55 inputting the data reads the contents of the buffers 56 and 57 and performs interpolation. The speech synthesis circuit 60f-j synthesizes speech using the smoothly combined speech parameters, and the speech output circuit 70 outputs the speech. Further, the next boundary between rIJ and [8AJ is also connected in the same way. However, at the boundary between ``sAJ'' and ``TU'', the first phoneme 1'-TJ of the following ``TUJ'' is a plosive, and its amplitude has a steep rise.
If interpolation is performed on the boundary between J and rIJ, its characteristics will fade. Therefore.

When combining the sound source amplitude parameters at the boundary between "SAJ" and "T tlJ, the speech parameter readout circuit 3゜b stops reading the sound source amplitude parameters and further ruptures the first phoneme ITJ of the following l'-T UJ. It is recognized as a sound,
The information is sent to the coupling control circuit 53b. Coupling control circuit 5
3b controls switch 51 and switch 58. Finally, the sound source amplitude parameter of the last part of the preceding l'-8AJ is inputted to the interpolation circuit 52b through the switch 51, and within the interpolation circuit 52b is inputted to the buffer 56 through the switch 58. Further, the combination control circuit 53b simultaneously controls the sound source amplitude parameter generation circuit 54 and the switch 59. The sound source amplitude parameter generation circuit 54 generates a sound source amplitude parameter whose sound source amplitude value is 0, and the switch 59 inputs the sound source amplitude parameter to the interpolator 55. Then, the interpolator 55 performs interpolation between the contents of the buffer 56 and the sound source amplitude parameter whose sound source amplitude value is 0 (the interpolation process for sound parameters other than the sound source amplitude parameter is performed at the boundary between rAJ and "work"). (same as processing)
.

With this interpolation process, the amplitude is once reduced to 0, and a waveform with a steep rise can be synthesized with the first sound source amplitude parameter of the next rTJ.

In addition, as an example of the present invention, the case where the following phoneme is a plosive has been explained, but the amplitude is set to 00 at the end of the sentence.
The present invention can also be used when setting the value to a constant value.

(Summary of the Invention) As explained above, the present invention provides a speech rule synthesis device for synthesizing arbitrary words, in which an amplitude with a steep rise is generated when combining speech units of synthesis. This is effective as a means for determining the amplitude of the sentence, and is also effective as a means for gradually reducing the amplitude at the end of a sentence.

[Brief explanation of drawings]

Figures 1 and 2 are block diagrams showing a conventional example, Figure 3 is a diagram showing an example of the waveform of a plosive in a sentence, and Figures 4 and 5 are block diagrams showing an example of the present invention. It is a figure. 10...Character code string input circuit, 20...
・Character code string-synthesis basic unit name string conversion circuit, 30a
. 30b...Audio parameter readout control circuit,
40...Audio parameter memory% soa
, s. b...Audio parameter combination circuit, 51...
...Audio parameter control switch, 52a, 52b...
...Interpolation circuit, 53a, 53b...Coupling control circuit, 54...Sound source amplitude parameter generation circuit, 55...Interpolator, 56.57... ...Buffer, 58.59...Switch, 60...
...Speech synthesis circuit, 70...Speech output circuit. -, knee/

Claims (1)

[Claims] means for inputting an arbitrary sentence as a character code string; means for converting the input character code string into a synthetic basic unit name string;
an audio parameter memory that stores in advance audio parameters (spectral parameters, sound source amplitude parameters, pitch parameters, etc.); means for reading audio parameters from the synthesis basic unit name sequence; and means for interpolating the audio parameters of the synthesis basic units. In a speech rule synthesis device comprising a coupling means for coupling and a means for synthesizing speech using the coupled speech parameters, the coupling means temporarily stores the sound source amplitude parameter of the final part of the preceding basic synthesis unit. a storage means 1, a storage means 2 for temporarily storing the sound source amplitude parameter of the first part of the subsequent basic synthesis unit, a constant value generation means for generating a predetermined sound source amplitude value, and a storage means 2 according to the name of the subsequent basic synthesis unit. and an interpolation means for interpolating the output of the selection means and the output of the storage means 1. A speech rule synthesizer that