JPH0594199A - Residual driving type speech synthesizing device - Google Patents

Abstract

PURPOSE:To obtain a composite voice of a more natural and high quality by constituting the device so that even if a residual waveform of any pitch period of a high-pitched sound, a middle-pitched sound and a low-pitched sound is used to a voice element piece selected in order to synthesize a voice, waveform data of all residual signals that can be selected is extracted by a reverse filter in which 2 voice parameter itself provided to actual synthesis is a coefficient, and a sound source and the parameter are extracted from the same voice in spite of a change of the pitch. CONSTITUTION:A voice parameter of a voice element piece unit of a voice element piece memory is used as a coefficient of a voice analytic filter consisting of a reverse filter of a voice synthesizing filter, waveform data of residual signals of a voice of three kinds of different pitch periods obtained by inputting a voice of pitch periods of a high-pitched sound, a middle-pitched sound and a low-pitched sound to this analytic filter, respectively are extracted, and these waveform data are accumulated in a residual waveform memory.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a regular speech synthesizer capable of uttering arbitrary words, and more particularly to a residual drive type regular speech synthesizer for performing residual drive.

[0002]

2. Description of the Related Art In recent years, research on a rule synthesizing method for synthesizing a voice from an arbitrary sentence has been actively conducted, and at present, a trial proofing device for a newspaper, a reading machine for blind persons, etc. have been put into practical use. is there.

A rule synthesizing device for synthesizing a voice from an arbitrary sentence, for example, performs a sentence analysis on a text input to determine a reading or an accent, and from a phonological rule,
A speech unit (for example, a CVC unit) that is a necessary synthesis unit is determined and combined, a voice pitch and the like are determined from a prosody rule, and a time series of voice parameters and a pitch pattern are generated. Synthesizing speech is generated by constructing a sound source and a digital filter from.

The speech parameters used in such a speech synthesis method are generally linear predictive parameters such as LPC and LSP, formants, and the like. On the other hand, as a sound source, memory reduction and simple processing are performed. Impulse and white noise were used for this purpose.

It is known that in a linear predictive speech synthesis method such as LPC or LSP, synthetic speech close to the original speech can be obtained by using the prediction residual as a driving sound source.

Therefore, even in the case of rule synthesis capable of uttering an arbitrary voice by inputting characters or the like, it is expected that a high-quality synthesized voice can be obtained from the above principle by inputting the residual as the driving sound source. Such a residual-driven type regular speech synthesizer has been proposed (Japanese Patent Application No. 2-2494).
No. 93), which will be outlined below.

FIG. 1 shows a configuration of a previously proposed residual-driven type regular speech synthesizer. According to this apparatus, when a character string to be uttered is input to a character string buffer (1), The phoneme symbol string generation unit (2) converts the input character string into a phoneme symbol string. For example, when a character string "ta * beniki * ta" (where * is a symbol indicating an accent position) is input, it is converted into a phoneme symbol string "tabenikita".

Next, the selection circuit 1 (3) sequentially determines and selects necessary speech units from the phoneme symbol string, and selects the necessary speech units among the speech units stored in the speech unit memory (4). One piece
It is connected by the voice unit connecting part (5). The unit of the speech unit is CVC (consonant + vowel + consonant) or CV
A combination of (consonant + vowel) and VC (vowel + consonant) is also used, but for the sake of simplicity, CV (consonant +
Considering vowels, that is, syllables as a unit, "t
"a, be, ni, ki, ta" are sequentially selected and connected as necessary speech units. The speech unit connected here is stored in the speech parameter buffer (6) as a speech parameter and given to the synthesis filter (11) as a coefficient.

On the other hand, a symbol representing intonation such as an accent position is also input together with the character string to be uttered,
When supplied from the character string buffer (1) to the pitch pattern generation unit (7), the pitch pattern generation unit (7)
Determines the pitch (pitch) of the entire utterance. For example,
In the case of inputting "ta * beniki * ta", "ta" and "ki"
Since there is an accent in the pitch pattern, the pitch pattern is as shown in FIG. In the pitch pattern generator (7),
A phrase component with a lower pitch and an accent component with a higher pitch at the accent position are added to the entire sentence,
A pitch pattern is generated.

Further, as shown in FIG. 3, a residual waveform for use as a driving sound source is stored in the residual waveform memory (8) corresponding to each speech unit, and the selection circuit 1
A required residual waveform is selected corresponding to the speech unit determined and selected in (3). In the case of the example, "ta, be, ni,
The corresponding residual waveforms are selected in the order of “ki, ta”. Further, as shown in FIG. 3, the residual waveform memory (8) also stores a plurality of residual waveforms having different pitches for each speech unit, and the residual waveform memory (8) generates them by the pitch pattern generation unit (7). The selection circuit 2 (9) selects and determines a residual waveform having an appropriate pitch according to the selected pitch, and the driving sound source generation unit (1
Store in 0). Finally, the selected residual waveform is subjected to pitch change in the driving sound source generation unit (10) to a value that matches the pitch generated in the pitch pattern generation unit (7), and the residual waveform of the desired pitch is obtained. Is generated.

The residual waveform thus generated is input to the synthesis filter (11) as a driving sound source, and synthetic speech is generated by the synthesis filter (11). Synthetic voice is D
The signal is output from the speaker (13) via the / A converter (12).

The operation of the driving sound source generation unit (10) of the already proposed device will be further described below.

First, FIG. 3 shows an example of waveform data of the residual signal stored in the residual waveform memory (8). The waveform data of the memory (8) corresponds to the case where a syllable having a CV (consonant + vowel) structure is used as a unit of a voice unit, and three types having different pitch periods are provided for each voice unit. , That is, the treble residual waveform (residual with a short pitch period), the middle tone residual waveform (medium residual pitch period),
The bass residual waveform (residue with a long pitch period) is stored. Then, the waveform data of the residual signal stored in the residual waveform memory (8) is processed as follows by the driving sound source generation unit (10) of FIG.

That is, the switch 1 (101) is selected by the selection signal from the selection circuit 1 (3) in FIG. 4, and the residual waveform corresponding to the necessary speech segment is stored in the residual waveform memory (8).
Are stored in the residual waveform buffer 1 (102). In the case of "Ta * Beniki * Ta", first, "ta"
The residual waveform of is read out. Next, the selection circuit 2 (9)
The switch 2 (103) is selected by the selection signal from (1) to select a residual with an appropriate pitch, and the residual is stored in the residual waveform buffer 2 (104). In the example, as shown in FIG. 2, the pitch of "ta" is as high as 400 Hz, so that the residual waveform of "ta" for the high tone is selected and the residual waveform buffer 2
It is stored in (104). Finally, the pitch changing circuit (105) changes the pitch cycle of the residual waveform so that the pitch is determined by the pitch pattern generator (7). For example, the residual waveform of "ta" for treble is 380H.
If it is extracted from the voice of z, it is set to 400 Hz determined by the pitch pattern generation unit (7).
A change is made so that the pitch becomes lower by Hz (pitch period becomes longer).

As the pitch change of the residual waveform performed by the pitch change circuit (105), for example, when lowering the pitch, zero data is inserted in the middle to lengthen the pitch cycle and increase the pitch. When this is done, the “zero-padded truncation method” is used in which data in the middle is deleted to shorten the pitch cycle.

If the pitch is changed drastically, the sound quality will be deteriorated. Therefore, in the above example of the residual drive type rule synthesizer,
By storing residuals with different pitches of three steps for high tone, middle tone, and low tone in the residual waveform memory in advance, and using the residual waveform having a pitch cycle close to the desired pitch, the amount of pitch change can be reduced. It is devised so that it can be reduced.

As described above, since the residual waveform corresponding to the selected speech unit and changed to the desired pitch is generated as the driving sound source, a synthetic speech with a high natural sounding can be obtained. Of.

According to the residual drive type speech synthesizer outlined above, the residual waveforms having different pitches stored in the residual waveform memory (8) are conventionally created by the method shown in FIG. Was common.

That is, for example, when creating residual waveforms of three types (three stages) of pitch, as shown in FIG.
By analyzing the treble input voice, the residual waveform for the treble with a short pitch period is extracted. Also, FIG.
As shown in (c), a low-pitched residual voice having a long pitch period is extracted by analyzing the low-pitched input voice. As for the middle tone, as shown in FIG.
By analyzing the input sound of middle tone, the residual waveform for the low tone having the average length of the pitch period is extracted.

[0020]

As described above, in the conventional residual drive type rule synthesizer, when the drive sound source is generated by using the residual waveforms having different pitches, it is stored as the voice unit. Since the parameters such as LPC and LSP and the residual waveform used as the driving sound source are those which are analyzed and extracted from different voices, the accumulated residual waveform is used as the driving sound source, and the accumulated speech unit is stored. Even if the parameter is used as a coefficient to pass through the synthesis filter, the original voice cannot be reproduced and the generated synthesized voice deteriorates.

Further, it is considered that the deterioration is eliminated by using the speech element extracted from the speech of the corresponding pitch for each residual waveform, but in this case, the speech element is corresponded to each pitch. It becomes necessary to store a plurality of them, which increases the amount of memory.

[0022]

According to the residual drive type speech synthesizer of the present invention, a speech parameter in a speech unit unit of a speech unit memory is used as a coefficient of a speech analysis filter which is an inverse filter of a speech synthesis filter. Waveform data of residual signals of voices having different pitch periods, which are obtained by inputting voices having different pitch periods to the analysis filter, and the waveform data is stored in a residual waveform memory.

[0023]

According to the residual drive type speech synthesizer of the present invention, no matter what period the residual waveform is used as the driving sound source for the speech unit selected for synthesizing the speech,
The waveform data of all the residual signals that can be selected is extracted by an inverse filter that uses the speech parameters themselves that are actually used for synthesis as coefficients, so that the parameters are the same as the sound source regardless of the pitch change. Since it is extracted from the voice, a more natural and high quality synthetic voice can be obtained.

[0024]

FIG. 6 shows a method of creating waveform data of a residual signal used in the residual drive type speech synthesizer of the present invention. Each H (z) in FIGS. 6A, 6B, and 6C represents the transfer characteristics of the voice analysis filter, which is an inverse filter of the voice synthesis filter, in the description of the Zet conversion. As shown in FIGS. 6 (a), 6 (b) and 6 (c), LPCs (linear prediction coefficients
By applying an inverse filter using voice parameters such as volume autocorrelation coefficient) and LSP as coefficients, residual waveforms for treble, middle tone, and bass are generated, respectively.

The residual drive type speech synthesizer of the present invention stores the residual waveform created by the same method as shown in FIG. 6 for each speech unit in the residual waveform memory shown in FIG. Residual drive type rule composition is performed with the configuration shown in FIG.

That is, when the character string to be uttered is input to the character string buffer (1), the phoneme symbol string generator (2) converts the input character string into a phoneme symbol string. For example,
When the character string "ta * beniki * ta" (where * is a symbol indicating an accent position) is input, "tabeni k
It is converted into a phoneme symbol string "ita".

Next, the selection circuit 1 (3) sequentially determines and selects necessary speech units from the phoneme symbol string, and selects the necessary speech units among the speech units stored in the speech unit memory (4). One piece
It is connected by the voice unit connecting part (5). The unit of the speech unit is CVC (consonant + vowel + consonant) or CV
A combination of (consonant + vowel) and VC (vowel + consonant) is also used, but for the sake of simplicity, CV (consonant +
Considering vowels, that is, syllables as a unit, "t
"a, be, ni, ki, ta" are sequentially selected and connected as necessary speech units.

The connected voice unit is stored in the voice parameter buffer (6) as a voice parameter and given to the synthesis filter (11) as a coefficient.

On the other hand, a symbol representing intonation such as accent position is also input together with the character string to be uttered,
When supplied from the character string buffer (1) to the pitch pattern generation unit (7), the pitch pattern generation unit (7)
Determines the pitch (pitch) of the entire utterance. For example,
In the case of inputting “ta * beniki * ta”, the pitch pattern is as shown in FIG. 5 because accents exist on “ta” and “ki”.

The residual waveform memory (8) stores residual waveforms to be used as a driving sound source corresponding to each voice segment, and is selected and selected by the selection circuit 1 (3). The required residual waveform is selected according to the speech unit.
In the case of the example, the residual waveforms are selected in the order of “ta, be, ni, ki, ta”.

Further, the residual waveform memory (8) stores a plurality of residual waveforms with different pitches created by the method proposed by the present invention for each speech unit, and the pitch pattern generator In accordance with the pitch generated in (7), the selection circuit 2 (9) selects and determines the residual waveform having an appropriate pitch,
It is stored in the driving sound source generator (10). Finally, the selected residual waveform is subjected to pitch change in the driving sound source generation unit (10) to a value that matches the pitch generated in the pitch pattern generation unit (7), and the residual waveform of the desired pitch is obtained. Is generated.

The residual waveform thus generated is input to the synthesis filter (11) as a driving sound source, and synthetic speech is generated by the synthesis filter (11). Synthetic voice is D
The signal is output from the speaker (13) via the / A converter (12).

[0033]

As described above, the residual drive type speech synthesizer of the present invention is selected according to the speech unit to be connected since the sound source and the parameter are extracted from the same speech regardless of the pitch change. With such a residual waveform, it is possible to obtain a high-quality synthesized speech close to the original speech by using the residual waveform of any pitch as the driving sound source.

[Brief description of drawings]

FIG. 1 is a block diagram of a residual drive type rule synthesizer,

FIG. 2 is a schematic diagram of a pitch pattern,

FIG. 3 is a schematic diagram of a residual waveform memory,

FIG. 4 is a schematic diagram of a driving sound source generation unit,

FIG. 5 is an explanatory diagram of a conventional residual waveform generation method,

FIG. 6 is an explanatory diagram of a residual waveform creating method used in the residual drive type speech synthesizer of the present invention.

[Explanation of symbols]

 (1) ... Character string buffer, (2) ... Phonological symbol string generator, (3) ... Selection circuit 1, (4) ... Speech segment memory, (5) ... Speech Unit connection unit, (6) ... Voice parameter buffer, (7) ... Pitch pattern generation unit, (8) ... Residual waveform memory, (9) ... Selection circuit 2, (10) .. drive sound source generation unit, (11), synthesis filter, (12), D / A converter, (13), speaker, (101), switch 1, (102), residual waveform buffer 1, (103) -switch 2, (104) -residual waveform buffer 2, (105) -pitch changing circuit

Claims (1)

[Claims] 1. A speech unit memory in which speech feature parameters of a linear prediction system such as LPC and LSP are stored in units of speech units, and a phonological symbol sequence generation unit for generating a symbol sequence indicating speech units of speech to be uttered. , And a voice feature parameter control means including a voice unit connection unit that sequentially connects the voice units read from the voice unit memory based on the symbol sequence generated by the phoneme symbol sequence generation unit, for each voice unit. Residual waveform memory that stores waveform data groups of a plurality of residual signals having different pitch periods corresponding to the above, a pitch pattern generation unit that generates a pitch pattern indicating a pitch period change of speech to be uttered, and the residual waveform A residual selection circuit that selects residual waveform data corresponding to the pitch cycle at each time point determined by the pitch pattern generation unit from the waveform data group of the residual signal corresponding to the speech unit of the memory. A driving sound source control means provided with the residuals selected by the residual selection circuit as a driving sound source, and a linear for synthesizing a voice by using a voice parameter of a voice unit connected by the voice unit connecting unit as a coefficient. In a residual drive type speech synthesizer comprising a speech synthesizer having a predictive speech synthesis filter, waveform data of a plurality of residual signals with different pitch periods to be stored in the residual waveform memory corresponds to this. As a coefficient, the voice parameter of the voice unit of the voice unit memory
A residual drive type speech synthesizer characterized by being obtained by inputting speech of different pitch periods to a speech analysis filter composed of an inverse filter of the speech synthesis filter.