JP2002358091A - Speech synthesis method and speech synthesis device - Google Patents

Abstract

(57) [Summary] [PROBLEMS] To provide a speech synthesis method and a speech synthesis device capable of naturally converting sentences into synthesized speech even when the input buffer size is limited. SOLUTION: A search is made for punctuation marks, end positions of parentheses, special symbols indicating emotions, and dependency information from input text data that are equal to or smaller than the input buffer size (steps S1 to S6). A character string smaller than the input buffer size is separated from the text data input based on the investigation result, and a combined unit character string is generated (step S7). Next, a synthesized speech is generated based on the generated synthesized unit character string, and a series of processing is repeated until all of the character strings of the text thus input are converted into voice, and all the character strings are converted into voice. (Step S8, S
9).

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice synthesizing method and a voice synthesizing apparatus for outputting voice based on an input character string.

[0002]

2. Description of the Related Art As a method of synthesizing a voice based on an input character string, there is a method of converting a character string into an emotional voice or a non-emotional voice. There is one as described in JP-A-7-72900.

In this method, character codes are parsed, accents of words and phrases are determined with reference to a dictionary constructed in advance, and various phonemes similarly constructed are expressed. Select the most appropriate one to compose the corresponding word from the characteristic parameters of the spectrum that can be obtained, further determine the intonation and power change of the whole sentence by rules, and output speech based on these. I have to.

[0004] In this technique, there is no particular limitation on the input buffer size of the synthesized speech.
Actually, a redundant buffer size was set so that the input character string would not exceed the buffer size.

[0005]

However, in such a conventional speech synthesis method, if the input buffer size is limited due to the memory size, a character string having the number of characters larger than the input buffer size is synthesized. There was a problem that it was not possible.

SUMMARY OF THE INVENTION The present invention has been made to solve the conventional problem, and provides a speech synthesis method and a speech synthesis apparatus capable of naturally converting sentences into synthesized speech even when the input buffer size is limited. Is what you do.

[0007]

SUMMARY OF THE INVENTION A speech synthesis method according to the present invention detects a punctuation mark, an end position of parentheses, a special symbol position indicating emotion, and dependency information from an input character string. Then, a synthesized unit character string having a length equal to or smaller than the input buffer size is separated from the character string, and a synthesized speech is generated for each separated synthesized unit character string.

According to such a method, even when a character string having the number of characters equal to or larger than the input buffer size is input, the sentence can be divided at a natural position, and a synthesized speech according to the sentence can be generated. it can.

Further, in the speech synthesis method according to the present invention, the special symbol indicating the emotion in the separated synthesized unit character string is sent from the emotion-added speech unit database and the reading-speech-based speech unit database. The speech unit database is selected to generate a synthesized speech.

According to such a method, even when a character string having the number of characters equal to or larger than the input buffer size is input, the sentence can be naturally separated, and a synthesized speech can be generated according to the sentence. In addition, it is possible to generate a synthesized speech to which an emotion corresponding to the special symbol indicating the emotion is added.

In the speech synthesis method according to the present invention, a character string enclosed in parentheses in the separated synthesis unit character string may be selected from a spoken speech unit database and a reading speech unit database. This is to generate a synthesized speech by selecting colloquial speech unit data.

According to such a method, even when a character string having the number of characters equal to or larger than the input buffer size is input, the sentence can be naturally separated, and a synthesized speech can be generated according to the sentence. , A character string enclosed in parentheses can generate a spoken synthesized voice.

Further, the speech synthesizing apparatus of the present invention includes a text input unit for inputting a text, a period search unit for searching for a period having a number of characters smaller than the input buffer size, and a search for a reading point having a number of characters smaller than the input buffer size. Reading point searching means, a parenthesis end position searching means for searching for an end position of a parenthesis having a number of characters equal to or less than the input buffer size, and a special symbol search indicating an emotion for searching for a special symbol indicating an emotion having a number of characters equal to or less than the input buffer size Means, dependency information checking means for checking dependency information of a character string having a number of characters equal to or smaller than the input buffer size, the period search means, the reading point search means, the parenthesis end position search means, the special symbol indicating the emotion A search unit configured to convert a text data character string input based on the A synthesizing unit character string generating means for separating a character string and generating a synthetic speech unit character string, a speech unit database for generating a synthetic speech, and synthesizing a character string generated by the synthesizing unit character string generating means It comprises a synthesized voice generating means for converting the voice data into voice data, and a voice output means for voice outputting the voice signal generated by the synthesized voice generating means.

With such a configuration, even when a character string having a number of characters equal to or larger than the input buffer size is input, a sentence can be naturally separated, and a synthesized speech can be generated according to the sentence. A synthesizer can be provided.

Further, the speech synthesizing apparatus of the present invention retrieves an emotion-attached speech unit database to which an emotion is attached, a reading-speech-based speech unit database, and a character string generated by the synthesis unit character string generation unit. When there is a special symbol indicating an emotion in the character string, the emotion-added speech unit database is selected. When there is no special symbol indicating the emotion in the character string, the reading-speech-based speech unit is selected. And a speech unit database selecting means for selecting a database.

With such a configuration, even when a character string having the number of characters equal to or larger than the input buffer size is input, the sentence can be naturally divided, and a synthesized speech in accordance with the sentence can be generated. In addition, it is possible to provide a speech synthesizer capable of generating a synthesized speech to which an emotion corresponding to a special symbol indicating an emotion is given.

Further, the speech synthesizer of the present invention searches a spoken speech unit database, a reading-speech unit database, and a character string generated by the synthesizing unit character string generating means, and searches for the character string. When there is a character string enclosed in parentheses, the colloquial speech unit database is selected, and when there is no character string enclosed in parentheses in the character string, the reading-speech-based speech unit database is selected. And a speech unit database selecting means for selecting

With such a configuration, even when a character string having a number of characters equal to or larger than the input buffer size is input, the sentence can be naturally separated, and a synthesized speech can be generated according to the sentence. , A character string enclosed in parentheses can provide a speech synthesizer capable of generating a spoken synthesized speech.

[0019]

Embodiments of the present invention will be described below with reference to the drawings.

FIGS. 1 and 2 show a first embodiment of a speech synthesis apparatus and a speech synthesis method according to the present invention.

First, the configuration will be described. In FIG. 1, the speech synthesizer of this embodiment includes a text input unit 1, a period search unit 2, a reading point search unit 3, a parenthesis end position search unit 4, a special symbol search unit 5 indicating emotion, and a dependency information search unit 6. , A synthesis unit character string generation means 7, a speech synthesis generation means 8, a speech unit database 9, and a speech output means 10.

A text document is input to the text input unit 1. The period search unit 2 searches the input text document for a period having a character number equal to or smaller than the input buffer size. The reading point searching means 3 searches for a reading point having a number of characters equal to or smaller than the input buffer size, and the parenthesis ending position searching means 4 searches for a parenthesis ending position having a number of characters equal to or smaller than the input buffer size.

Further, the special symbol search means 5 is designed to search for a special symbol indicating an emotion having the number of characters less than the input buffer size, and the dependency information search means 6 is used for the dependency of a character string smaller than the input buffer size. They are looking for information.

The information obtained by the combining unit character string generation means 7 by the period search means 2, the reading point search means 3, the parenthesis end position search means 4, the special symbol search means 5 indicating emotion, and the dependency information search means 6 , A character string smaller than the input buffer size is separated from the text data character string, and a combined unit character string is generated. The synthesized speech generation means 8 generates a synthesized speech of the character string generated by the synthesis unit character string generation means 7 using the speech unit database 9.

The voice output means 10 outputs the synthesized voice generated by the synthesized voice generation means 8.

Next, a speech synthesis method will be described with reference to the flowchart shown in FIG.

First, when a text document input process is performed (step S1), a process for searching for a period having a number of characters equal to or smaller than the input buffer size is executed by the period searching means 2 (step S2).

Next, after the reading point searching means 3 performs a process of searching for a reading point having a number of characters smaller than the input buffer size (step S3), the parenthesis ending position searching means 4 ends the parentheses having the number of characters smaller than the input buffer size. A process for searching for a position is executed (step S4). Next, the special symbol searching means 5 executes a process of searching for a special symbol indicating an emotion having the number of characters equal to or smaller than the input buffer size (step S5). A process for checking the received information is executed (step S6).

Next, after the processing for generating a synthesized unit character string is executed by the synthesized unit character string generating means 7 (step S7), the processing for generating synthesized speech is executed by the speech synthesis generating means 8 (step S8). Next, it is determined whether or not all the input character strings have been converted to voice (step S9). If the character strings have been converted, the current process is terminated. If not, the process returns to step S2. Then, a series of processes is repeated until all the character strings of the input text are converted into voice by executing the processing after step S2, and all the character strings are converted into voice.

As described above, according to the present embodiment, even when a character string having a number of characters equal to or larger than the input buffer size is input, a sentence can be naturally separated and a synthesized speech can be generated according to the sentence.

FIGS. 3 and 4 are diagrams showing a second embodiment of the voice synthesizing method and the voice synthesizing apparatus according to the present invention. In the present embodiment, the same configuration as the first embodiment includes:
The same numbers are assigned and the description is omitted.

The second embodiment differs from the first embodiment in that it has a speech unit database selection means 11, a reading-speech-based speech unit database 12, and an emotion-added speech unit database 13, as shown in FIG. It is.

The speech unit database selecting means 11 determines whether or not the synthesized unit character string generated by the synthesized unit character string generating means 7 includes a special symbol indicating an emotion. Is selected, the emotion-attached speech unit database 13 is selected. Otherwise, the reading-speech-based speech unit database 12 is selected.
Is to be selected.

In addition, the emotion-added speech unit database 13 stores a segment to which an emotion is added, and the reading-speech-based speech unit database 12 stores a reading-tone unit to which no emotion is added. ing.

Next, a speech synthesis method will be described with reference to FIG.

When the input process of the text document is performed (step S11), a process of searching for a period having the number of characters equal to or smaller than the input buffer size is executed by the period searching means 2 (step S12).

Next, after the reading point searching means 3 performs a process of searching for a reading point having a number of characters smaller than the input buffer size (step S13), the parenthesis ending position searching means 4 ends the parentheses having the number of characters smaller than the input buffer size. A process for searching for a position is executed (step S14). Next, the special symbol searching means 5 performs a process of searching for a special symbol indicating an emotion having the number of characters equal to or smaller than the input buffer size (step S15). A process for checking the received information is executed (step S16).

Next, after the synthesis unit character string generation means 7 executes processing for generating a synthesis unit character string (step S17), the speech unit database selection means 11 generates the synthesis unit character string generation means 7. It is determined whether or not the unit character string includes a special symbol indicating an emotion.If the unit character string includes a special symbol indicating an emotion, the unit 13 selects the emotion-giving speech unit database 13, and if not, otherwise. Select the reading voice unit database 12 (step S1)
8).

Next, after executing a process of generating a synthesized speech of the database selected by the speech synthesis generating means 8 (step S19), it is determined whether or not all the input character strings have been converted into speech (step S19). S20) If the conversion has been performed, the current processing is terminated. If the conversion has not been performed, the process returns to step S12 to execute the processing of step S12 and thereafter, and all the character strings of the input text are converted to voice. A series of processing is repeated until conversion, and all character strings are converted to speech.

As described above, in the present embodiment, even when a character string having the number of characters equal to or larger than the input buffer size is input, the sentence can be naturally separated, and synthesized speech can be generated according to the sentence. It is possible to generate a synthetic voice to which an emotion is given according to the special symbol indicating the emotion.

FIGS. 5 and 6 show a third embodiment of the voice synthesizing method and the voice synthesizing apparatus according to the present invention. In the present embodiment, the same configuration as the first embodiment includes:
The same numbers are assigned and the description is omitted.

The present embodiment differs from the first embodiment in that it has a speech unit database selecting means 21, a reading-speech-based speech unit database 22, and a spoken-speech-based speech unit database 23, as shown in FIG. It is.

The speech unit database selection means 11 is adapted to determine whether or not the synthesis unit character string generated by the synthesis unit character string generation means 7 includes a character string enclosed in parentheses. If there is a character enclosed in parentheses, the colloquial speech unit database 23 is selected, and if there is no character enclosed in parentheses, the reading unit 22 is selected. I have.

Further, the reading-speech-speech unit database 22 stores speech-segment segments to which no emotion is given, and the spoken-speech unit database 23 stores speech-speech units. .

Next, a speech synthesis method will be described with reference to FIG.

When the input process of the text document is performed (step S21), a process of searching for a period having the number of characters smaller than the input buffer size is executed by the period searching means 2 (step S22).

Next, after the reading point searching means 3 performs a process of searching for a reading point having a number of characters smaller than the input buffer size (step S23), the parenthesis ending position searching means 4 ends the parentheses having the number of characters smaller than the input buffer size. A process for searching for a position is executed (step S24). Next, the special symbol searching means 5 executes a process of searching for a special symbol indicating an emotion having the number of characters equal to or smaller than the input buffer size (step S25). A process for checking the received information is executed (step S26).

Next, after performing a process of generating a synthesized unit character string by the synthesized unit character string generating unit 7 (step S27), the synthesis unit character string generating unit 7 generates the synthesized unit character string by the speech unit database selecting unit 11. Determines whether the unit string contains a character string enclosed in parentheses,
If a character string enclosed in parentheses is included, the spoken voice unit data 23 is selected, otherwise, the reading voice unit database 22 is selected (step S28).

Next, after executing a process of generating a synthesized speech of the database selected by the speech synthesis generating means 8 (step S29), it is determined whether or not all the input character strings have been converted into speech (step S29). S30) If the conversion has been performed, the current processing is terminated. If the conversion has not been performed, the process returns to step S22 to execute the processing of step S22 and thereafter, and all the character strings of the input text are converted to voice. A series of processing is repeated until conversion, and all character strings are converted to speech.

As described above, in the present embodiment, even when a character string having the number of characters equal to or larger than the input buffer size is input, it is possible to naturally separate sentences and generate synthesized speech according to the meaning of the sentence. A character string enclosed in parentheses can generate a spoken synthesized voice.

[0051]

According to the present invention, a character string input so as to have a number of characters equal to or smaller than the input buffer size is separated as a synthetic unit character string, and is converted into a synthetic voice for each synthetic unit character string. There is an effect that a sentence can be naturally converted into a synthesized speech even when the buffer size is limited.

[Brief description of the drawings]

FIG. 1 is a diagram showing a first embodiment of a speech synthesis method and a speech synthesis device according to the present invention, and is a block diagram of the speech synthesis device;

FIG. 2 is a flowchart of a speech synthesis method according to the first embodiment;

FIG. 3 is a diagram showing a second embodiment of the speech synthesis method and the speech synthesis device according to the present invention, and is a block diagram of the speech synthesis device;

FIG. 4 is a flowchart of a speech synthesis method according to a second embodiment.

FIG. 5 is a diagram showing a third embodiment of the speech synthesis method and the speech synthesis device according to the present invention, and is a block diagram of the speech synthesis device;

FIG. 6 is a flowchart of a speech synthesis method according to a third embodiment.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Text input means 2 Term search means 3 Reading point search means 4 Parentheses end position search means 5 Special symbol search means showing emotion 6 Dependency information search means 7 Synthesis unit character string generation means 8 Speech synthesis generation means 9 Speech unit database 10 Speech output means 11 Database selection means 12 Reading-speech speech unit database 13 Emotional speech unit database 21 Database selection means 22 Reading-speech speech unit database 23 Spoken speech unit database

Claims (6)

[Claims] 1. A punctuation mark, an end position of parentheses, a position of a special symbol indicating an emotion, and dependency information having a number of characters equal to or less than an input buffer size are detected from an input character string. A speech synthesis method characterized in that the synthesis unit character string is separated and a synthesized speech is generated for each of the separated synthesis unit character strings. 2. A special symbol indicating an emotion in the separated synthesized unit character string is selected from an emotion-attached speech unit database from an emotion-attached speech unit database and a reading-speech-based speech unit database. 2. The speech synthesis method according to claim 1, wherein a synthesized speech is generated. 3. For the character string enclosed in parentheses in the separated synthesized unit character string, colloquial speech unit data is selected from a colloquial speech unit database and a reading-speech unit speech database. 3. The speech synthesis method according to claim 1, wherein the speech synthesis method generates a synthesized speech. 4. A text input unit for inputting a text, a period search unit for searching for a period having a number of characters equal to or smaller than the input buffer size, a reading point searching unit for searching for a period having a number of characters equal to or smaller than the input buffer size, and an input buffer. A parenthesis end position search means for searching for an end position of a parenthesis having a number of characters equal to or less than the size, a special symbol searching means for indicating an emotion for searching for a special symbol indicating an emotion having a number of characters equal to or less than the input buffer size, Dependency information investigating means for investigating dependency information of a character string which is the number of characters, the period search means, the reading point search means, the parenthesis end position search means, the special symbol search means indicating the emotion, the dependency information search A character string smaller than the input buffer size is separated from the text data character string input based on the , A speech unit database for generating a synthesized speech, and a synthesized speech generation means for converting the character string generated by the synthesized unit character string generation means into synthesized speech data, A voice output unit that outputs a voice signal generated by the synthesized voice generation unit. 5. Searching an emotion-attached speech unit database to which an emotion has been added, a reading-speech-based speech unit database, and a character string generated by the synthetic unit character string generation means,
When there is a special symbol indicating an emotion in the character string, the emotion-added speech unit database is selected. When there is no special symbol indicating an emotion in the character string, the reading-speech-based speech unit database is selected. 5. The speech synthesis apparatus according to claim 4, further comprising: a speech unit database selection unit for selecting a speech unit. 6. A colloquial speech unit database, a reading-speech unit database, and a character string generated by the synthesis unit character string generation unit, and a character string enclosed in parentheses is included in the character string. Speech unit database selection for selecting the spoken speech unit database when there is a sequence, and for selecting the reading tone speech unit database when there is no character string enclosed in parentheses in the character string. The voice synthesizing apparatus according to claim 4 or 5, further comprising: