JPS61166600A - Voice snthesizer - 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] (b) Industrial application fields The present invention relates to a speech synthesis device that synthesizes speech.

(B) Prior art As proposed in Japanese Patent Application No. 59-2291, a conventional speech synthesis device has a filtering section and a regular vowel section in the case of the Japanese language, where the consonant section is passed from the consonant section to the vowel section. LSP coefficients corresponding to approximately 100 connected dorsal segments, PARO
There is a system that stores syllable parameters such as OR coefficients in a ROM (read only memory) and resynthesizes speech in units of words and sentences by connecting the syllable parameters read from the ROM. This synthesis method is generally known as a rule synthesis method.

Such a speech synthesizer can resynthesize any Japanese speech by storing syllable parameters for approximately 100 syllables in its memory, but the output speech is based on syllable data stored in the ROM. Therefore, it has the disadvantage that it can only perform resynthesis of voices that have those characteristics. Therefore, it has been impossible to output synthesized speech that has the characteristics of various people.

(c) Problems to be Solved by the Invention The present invention has been made in view of the above-mentioned points, and provides a speech synthesis device having a rule synthesis method.

B) Means for Solving the Problems The speech synthesis device of the present invention has a feature of inputting speech from the outside and extracting characteristic parameters indicating individuality.

(e) Operation The speech synthesis device of the present invention stores syllable parameters in an iROM (read-only memory) in which a consonant part, a filter transition part from the consonant part to the consonant part, and a constant vowel part are combined. In addition, the vowel parameters indicating stationary individuality are input from the back door input unit and the speech analysis unit, and stored in the buffer memory, and the consonant part extracted from the stationary vowel parameter and the syllable parameter above, and the vowel part from the consonant part are input. A syllable parameter generation unit that connects Tsuguri and filtered Watabe to 11 and 1 and uses it as a syllable parameter to a new syllable unit with 1 personal information added, and is it? The syllable parameter buffer memory is used to synthesize speech using the above syllable parameter string and pitch parameters that match the syllable parameter string. (f) Embodiment FIG. 1 shows an embodiment of the speech synthesis apparatus of the present invention. In the same figure, (1 (l is the back door input microphone.

(11 is a parameter extraction circuit that analyzes the steady vowel input from the voice input microphone α and extracts voice characteristic parameters such as LSF parameters and PAROOR parameters. (121 is a parameter extraction circuit that This is a stationary vowel buffer memory (RAM) that temporarily stores characteristic parameters of stationary vowels.

Further, (4) is a standard syllable parameter memory in which various syllable parameters in which a consonant part, a transition part that continues from the consonant part to a vowel part, and a constant vowel part are combined are stored with addresses. . (6) is the syllable parameter in the standard syllable parameter memory (4)! .

Extract only the shiji dorsal part and the consonant part power/ra vowel, and combine the extracted data with the stationary vowel in the above stationary vowel buffer memory [17J and create a new (consonant part) +
This is a syllable parameter generation unit that generates a new syllable parameter having the personality information of the person who inputs the steady vowel: (transient part) + (input steady vowel). (5) is a user syllable parameter memory (R
AM).

On the other hand, (1) is a keyboard with character keys arranged, (2)
is a decoder that receives key operation signals from the keyboard (1) and converts them into character signals in syllable units corresponding to the keys. (3) is a syllable address table that links the character signal from the decoder (2) with each syllable address in the standard syllable parameter memory (4) and user syllable parameter memo (January 5). Further, (161) is a syllable length table in which the character signal in syllable units from the decoder (2) is associated with the generation time length of that syllable.

Further, (7) is a synthesized voice selection unit which selects which syllable parameter from the standard syllable parameter memo (January 4) or the user's syllable parameter memory (5) is used for speech synthesis. Also, (8) is the syllable length table (16
This is a syllable data length control unit that expands or compresses the syllable data length so as to match the syllable generation time length specified in 1. (9) is the parameter area (9-a) and the pitch area (9
-11) is an audio data buffer memory consisting of
In the parameter area (9-a), the standard syllable parameter memory (4) or the user syllable parameter memo (January 5) is stored as a syllable parameter in a state where it is interleaved by the syllable data length control unit (8). New syllable parameters are sequentially stored in accordance with the subsequent key presses from the keyboard (1).

α3 is an accent specification section for specifying the accent type of synthesized speech. α4 is a pitch pattern designation circuit that generates a pitch pattern designation signal consisting of a combination signal of the accent designated by the accent designation unit (13) and the number of moras represented by the number of syllables of the synthesized speech obtained by inputting the keyboard (1). be.

(Is is a pitch table containing a standard pitch back and meter that determines the intonation and transition of the synthesized voice from the pitch pattern designation signal from the pitch pattern designation circuit [141]).

Pitch parameters are stored in patterns for each combination of morph number and accent type.

That is, the pitch frequency of the accent position is set to be relatively high. Sum η is the syllable length table t above.
This is a pitch pattern matching circuit that linearly compresses or linearly expands the standard pitch pattern IH of the synthesized speech obtained from the pitch table 9 based on the time length of each syllable from t61. The matched pitch pattern is the audio data buffer memo 1
This matching pitch pattern and the parameter area (9-I) are stored in the pitch area (9-11) of
L) is associated with the syllable parameter string.

The bet is made by inputting the syllable parameter string stored in the audio data buffer memory (9) and the matching pitch pattern corresponding thereto, using the keyboard (1).
This is a voice synthesis unit that synthesizes and outputs voice signals corresponding to input. (11 is an amplifier that amplifies the synthesized voice output from the voice synthesizer, and the final synthesized voice is generated from the speaker ring.

Next, the processing procedure in the syllable parameter generation unit (6) is #
A detailed explanation will be given based on the flowchart shown in Figure IIz.

First, the parameter is extracted from the standard syllable parameter memory (4), and it is determined whether the parameter is a consonant part and whether the transition from the consonant part to the consonant part is a filtered part or a consonant part. If the transition part is a transition part, write it into the user syllable parameter memory, and do this until the end of the data of the continuation part to a stationary consonant part.Furthermore, if the parameter from the standard syllable parameter memory (4) is a vowel, Instead of this parameter, the stationary vowel buffer memory +13 imports the corresponding stationary vowel parameter.

0 in the user syllable parameter memory as shown in the SXS diagram
pair 2s 7' (consonant part → ゛(stationary consonant part'') continued
Continuing to write the data in part 1). This work, + standard syllable puff meter memo January 4
) for all syllable buff meters. As a result, a syllable parameter having new individuality information corresponding to the standard syllable parameter memory (4) is generated in the needle all syllable parameter memo (January 5).

(E) Effects of the Invention As is clear from the above description, the speech synthesis device of the present invention includes a feature parameter extraction section that inputs voice from the outside and extracts feature parameters indicating individuality, and a feature parameter extraction section that extracts feature parameters indicating personality. Since the system is equipped with a syllable parameter generation unit that generates syllable parameters to which personal information is added using the characteristic parameters of the user, it is possible to perform rule synthesis using syllable parameters that include the user's personal information as a basic unit. It is possible to output a synthesized voice that is close to the user's own voice.

[Brief explanation of drawings]

Figure @1 is a block diagram showing the configuration of an embodiment of the speech synthesis device of the present invention, and Figures 2 and '@3 are flowcharts showing the processing procedure of the syllable parameter generation section according to the device of the present invention, and their flowcharts. It is a memory diagram. (1)... Keyboard, (2)... Decoder, (3
)...Syllable address table, (4)...Standard syllable parameter memory. (5)...User syllable parameter memory, (6)...
...Syllable parameter generation unit, (7)...Synthesized speech selection unit, (8)...Syllable data length control unit, (9)...Speech data backup memory. (9-&)...Parameter area, (9-7))...
・Pitch area, voice input microphone, α parameter extraction circuit, α constant vowel buffer memory, G
Accent specification section, α4...Pitch pattern specification circuit. 15... Pitch table, a = node length table, (1
η...Pitch pattern matching circuit, Q81"-
fVoice synthesis section, (1!l...Amplifier, ■...Speaker.

Claims (1)

[Claims] In a speech synthesis device that synthesizes arbitrary word or sentence speech based on each syllable parameter formed by connecting syllable parameters, which are feature parameters of speech in syllable units, feature parameters indicating individuality are obtained by inputting speech from an external source. and a syllable parameter generation unit that uses the feature parameters obtained by the extraction unit to generate syllable parameters to which new personal information is added. Synthesizer.