JPH03276194A - Text/sound converter - Google Patents

Abstract

PURPOSE:To obtain a desired synthesized tone equipped originally with reading, accent, intonation and breath, etc., which are originally controlled by a user, with simple configuration by providing mode set and control parts. CONSTITUTION:When the mode set part 40 selects any one of modes such as a text/ sound conversion mode, phoneme sound and meter symbol train output mode and phoneme sound and meter symbol train synthesizing mode according to a designation from an external part, the control part 50 discriminates the mode and controls the input/output of a text and a phoneme sound and meter symbol train. When the text/ sound conversion mode is set, the control part 50 inputs the text and a text analysis part 30 analyzes the text. Then, the result of the analysis is inparted through a sound synthesizing part 60 to a loudspeaker 61. When the phoneme sound and meter symbol train output mode is set, the control part 50 analyzes the text at the text analysis part 30 and outputs the generated pheneme sound and meter symbol train in the form of a character code to the external part. When the phoneme sound and meter symbol train synthesizing mode is set, the control part 50 directly outputs the phoneme sound and meter symbol train inputted from the external part, through the sound synthesizing part 60. Thus, the user can freely change the pheneme sound and meter symbol train and easily obtain the desired synthesized tone.

Description

[Detailed Description of the Invention] (Industrial Application Field) The present invention is a text-to-speech conversion system that inputs text data consisting of character codes (this is referred to as "text"), converts it into speech, and outputs it. It is related to the device.

(Prior art) Conventionally, as a technology in this field, Japanese Patent Application Laid-Open No. 1-119
There was one described in Publication No. 822.

Conventionally, speech synthesis devices include a text-to-speech conversion device that inputs character information, converts it into speech, and outputs it, and a speech synthesizer that stores speech in a memory or the like and plays it back. There were storage and playback type speech synthesis devices. The latter device generates only stored speech, so it is always limited in its output vocabulary, but the former device converts input text information into speech and outputs it, so there is no limit to the output vocabulary. Therefore, it can be expected to be applied in various fields as a technology to replace the latter device.

An example of this type of device is shown in FIG.

FIG. 2 is a block diagram showing a configuration example of a conventional text-to-speech conversion device described in the above-mentioned literature.

A text message connected to the output side of host computer 1.
The speech conversion device 10 includes an interface section 11, a text analysis section 12, a speech synthesis section 13, and a speaker 14.

In this device, when text is output from the host computer 1, the text is provided to the text analysis section 12 via the interface section 11. The text analysis unit 12 segments the input text into words,
Phonological and prosodic symbol strings necessary for reading out the text aloud, such as intonation symbols representing the pronunciation of each word, accent, and intonation of the entire text, are generated and sent to the speech synthesis unit 13. The speech synthesis section 13 synthesizes speech corresponding to the input sentence, that is, the input text, based on the phoneme and prosodic symbol strings, and outputs the synthesized speech from the speaker 14.

(Problems to be Solved by the Invention) However, the apparatus with the above configuration has the following problems.

In conventional text-to-speech conversion devices that read out sentences,
Since it has the advantages of simple configuration, small circuit scale, and low cost, it is used for various purposes.

However, at present, it is nearly impossible for this type of text-to-speech conversion device to analyze sentences that are used in daily life, including kanji and kana, and obtain perfect pronunciation, accent, and intonation. In addition, in order to obtain accurate phonological and prosodic symbol strings, very sophisticated analysis including semantic analysis is required.
This causes the problem of complicating the device configuration, thereby increasing the circuit scale and cost. Therefore, in conventional text-to-speech conversion devices that read sentences aloud, the phoneme and monotony symbol strings that are the result of text analysis often contain errors in word pronunciation, accent, or sentence intonation, making it difficult for the user to understand. It was difficult to obtain the expected synthesized sound. Furthermore, with conventional text-to-speech conversion devices that input text, automatically analyze it internally, and read it out as voice, it is impossible to finely adjust the synthesized sound as the user intends. We had to settle for ready-made synthesized sounds based on analysis by the authors, and we were still unable to obtain anything that was technically satisfactory.

The present invention addresses the problem of the prior art, which is that it is difficult for a user to obtain a desired synthesized voice with a unique reading, accent, intonation, breath pause, etc. with a simple configuration. The present invention provides a text/speech conversion device that solves the above problems.

(Means for Solving the Problems) In order to solve the problems described above, the present invention provides a text analysis unit that analyzes kanji-kana mixed text data and generates a phonological and prosodic symbol string for the text data, This text-to-speech conversion device includes a speech synthesis section that outputs synthesized speech corresponding to the text data based on a symbol string, and a mode setting section and a control section.

Here, the mode setting section includes a text-to-speech conversion mode in which the text data is input to the text analysis section, a phonology in which the phonology and prosodic symbol strings generated in the text analysis section are outputted to the outside in the form of character codes; It has a function of externally setting any one of a prosodic symbol string output mode and a phoneme and prosodic symbol string synthesis mode in which phoneme and prosodic symbol strings from the outside are directly input to the speech synthesis section. Further, the control section has a function of determining the mode state set by the mode setting section, and controlling input and output of the text data, phoneme, and prosodic symbol string based on the determination result.

(Function) According to the present invention, since the text-to-speech conversion device is configured as described above, the mode setting unit can select the text-to-speech conversion mode and the phonetic and prosodic symbol string output mode according to external specifications. When one of the phoneme and prosodic symbol string synthesis modes is selected, the control unit determines the set mode state, and inputs and outputs the text, phoneme, and prosodic symbol string based on the determination result. control.

For example, when the text-to-speech conversion mode is set, the control unit inputs text, provides it to the text analysis unit to analyze it, and causes the speech synthesis unit to output synthesized speech based on the analysis result. When the phonological and metrical symbol string output mode is set, the control section inputs text, causes the text analysis section to analyze the text, and converts the phonological and metrical symbol strings generated by the text analysis section into character code formats. output to the outside. Further, when the phoneme and prosodic symbol string synthesis mode is set, the control section directly inputs the phoneme and prosodic symbol string input from the outside to the speech synthesis section, and causes the speech synthesis section to output synthesized speech.

In this way, since the configuration is such that three modes can be selected by external specification, the user can freely change the phonological and prosodic symbol strings that are the result of text analysis, thereby making it possible to easily create the desired synthesized sound. It can be obtained with Therefore,
The above problem can be solved.

(Embodiment) FIG. 1 is a configuration block diagram of a text-to-speech conversion device showing an embodiment of the present invention.

This text-to-speech conversion device 20 is connected to the output side of a host computer 1 having arithmetic and control functions, and includes an interface unit 21 that sends and receives signals between the host computer 1 and the inside of the device, and a word reading,
A word dictionary memory 22 consisting of a ROM (read-only memory) etc. in which accent positions, parts of speech information, etc. are stored.
It is equipped with. A text analysis section 30 is connected to this word dictionary memory 22 .

The text analysis unit 30 analyzes the input text with reference to the contents of the word dictionary memory 22, and obtains phonemes necessary for synthesizing text data into speech, such as word pronunciation, accent position, and intonation marks representing sentence intonation. , which generates a prosodic symbol string. This text analysis section 30 includes word division processing means 31 that divides the text into words;
Reading assignment processing means 3 for assigning a reading to each divided word
2. Accent adding processing means 33 for adding an accent
, and a pause/intonation setting means 34 for setting the pause and intonation.

The text-to-speech conversion device 20 is also provided with a mode setting section 40, and a control section 50 is connected to the mode setting section 40 and the text analysis section 30. A speaker 6 is connected to the output side of the control unit 50 via a voice synthesis unit 60.
1 is connected.

The mode setting unit 40 has a text-to-speech conversion mode in which text is input to the text analysis unit 30, and a phonology mode in which the phoneme and prosodic symbol strings generated by the text analysis unit 30 are output to the external host computer 1 in the form of character codes. , a prosodic symbol string output mode, and a phonological and prosodic symbol string synthesis mode in which phonetic and prosodic symbol strings from the external host computer 1 are directly input to the speech synthesis section 60. It has the function to set. The mode setting section 40 has, for example, a changeover switch, and either manually switches the mode or transfers commands from the host computer 1 to the interface section 21.
It has a function of switching the mode by receiving the information via the control unit 50.

The control unit 50 includes a mode determination unit 51 that determines the mode state set by the mode setting unit 40, and an input/output selection unit 5.
It is equipped with 2. The input/output selection means 52 selects the interface section 21 based on the determination result of the mode determination means 51.
Input text, phoneme, and prosodic symbol string from , output the phoneme and prosodic symbol string generated by the text analysis section 30 to the interface section 21 , and roughly synthesize the output of the text analysis section 30 or the output of the interface section 21 into speech. Part 60
It has an input/output function to send data to.

The speech synthesis section 60 has a function of outputting synthesized speech corresponding to the text data to the speaker 60 based on the input phoneme and prosodic symbol strings.

Next, the operation will be explained.

For example, the following operation will be described assuming that the mode setting section 40 manually sets the mode in advance.

When the text-to-speech conversion mode is set by the mode setting section 40, the text-to-speech conversion device 20 converts the text from the host computer 1 to the interface section 2.
1 to the control unit 50. In the control section 50km, the mode determination means 51 determines the mode state in the mode setting section 40, and determines that it is the text/speech conversion mode. Then, based on the determination result, the input/output selection means 52 sends the text received from the interface section 21 to the text analysis section 30.

In the text analysis section 30, the word division processing means 31 refers to the contents of the word dictionary memory 22 and divides the input text into words. The reading assignment processing means 32 assigns a pronunciation to each of the divided words, the accent assignment processing means 33 assigns an accent, and the rough pause/intonation setting section 34 determines the pause and overall sentence. The intonation symbol representing the intonation is set, and a phoneme and @meter symbol string are generated and sent to the input/output selection means 52.

The input/output selection means 52 sends the phoneme and prosodic symbol string output from the text analysis section 30 to the speech synthesis section 60 based on the output of the mode determination means 51. The speech synthesis section 60 synthesizes speech based on the phoneme and prosodic symbol strings received, and outputs the synthesized speech from the speaker 61.

Next, the operation when the mode setting section 40 sets the phoneme and prosodic symbol string output mode will be described.

In the phonetic and prosodic symbol string output mode, text from the host computer 1 is input to the control section 50 via the interface section 21, as in the text-to-speech conversion mode. The mode determining means 51 in the control section 50 determines the mode state in the mode setting section 40, and when determining that it is the phoneme and prosodic symbol string output mode, provides the determination result to the input/output selecting means 52. Input/output selection means 52
sends the received text from the interface section 21 to the text analysis section 30. Then, the text analysis section 30
Now, the input text is analyzed in the same manner as described above to generate phoneme and prosodic symbol strings, which are sent to the input/output selection means 52.

The input/output selection means 52 selects phonemes expressed in the form of character codes obtained by the analysis of the text analysis section 30,
The prosodic symbol string is sent to the host computer 1 via the interface section 21.

In the host computer 1, by storing the phoneme and prosodic symbol strings received from the interface unit 21 in a file, they can be easily modified using, for example, a normal editor (a program for editing and modifying source programs).

Further, the operation when the mode setting unit 40 sets the phoneme and prosodic symbol string synthesis mode will be explained.

This phoneme and prosodic symbol string synthesis mode is a mode in which speech is synthesized using phoneme and prosodic symbol strings that have been modified with an editor or the like so that the user can obtain a desired synthesized sound, for example. In this mode, the phoneme and prosodic symbol strings output from the host computer 1 are input to the control section 50 via the interface section 21. The mode determining means 51 in the control section 50 determines the mote state in the mode setting section 40, and when determining that the mode is the phoneme/prosodic symbol string synthesis mode, provides the determination result to the input/output selecting means 52.

The input/output selection means 52 sends the phoneme and prosodic symbol strings received from the interface section 21 directly to the speech synthesis section 60 without passing through the text analysis section 30.
0 generates a synthesized sound.

This embodiment has the following advantages.

In this embodiment, in addition to the function of inputting normal text (sentence data containing kanji and kana) and converting it into speech, the phonological and prosodic symbol strings, which are intermediate results analyzed by the text analysis section 30, are converted into character codes. It has a function of outputting to the host computer 1 and a function of inputting phoneme and prosodic symbol strings from the host computer 1 and converting them directly into speech.
These three functions can be selected by the mode setting section 40 and the control section 50 according to external specifications. Therefore, it is possible to output synthesized speech based on the text analysis results by the device, as well as the phonology and phonology that are the text analysis results.
User-adjusted readings for prosodic symbol strings,
It is possible to obtain a desired synthesized sound with accent, intonation, pause, etc. Moreover, in the present invention, since it is only necessary to add the mode setting section 40 and the control section 50, the configuration of the device is simplified, and thereby the circuit size and cost can be reduced.

Note that the present invention is not limited to the illustrated embodiment, and various modifications are possible. For example, various modifications can be made, such as adding other processing means to the text analysis section 30 or the control section 50, or configuring them with individual circuits, or configuring them to be executed by a computer program III1 or the like. It is possible.

(Effects of the Invention) As described in detail above, according to the present invention, a mode setting unit is provided for setting a text-to-speech conversion mode, a phoneme, a prosodic symbol string output mode, a phoneme, a prosodic symbol string synthesis mode, Roughly, based on the output of the mode setting section, a control section is provided that controls the input and output of text data, phonemes, and prosodic symbol strings.

Therefore, by setting the text/speech conversion mode using the mode setting section, it is possible to input a normal sentence containing kanji and kana and convert it into speech. Furthermore, when the mode setting section sets the phoneme and prosodic symbol string output mode, the phoneme and prosodic symbol strings generated by the text analysis section are outputted to the outside via the control section in the form of a character code. This allows the user and the like to freely change the phoneme and prosodic symbol strings that are the text analysis results. After setting the phoneme and prosodic symbol string synthesis mode in the mode setting section, by inputting the phoneme and prosodic symbol strings changed by the user from the outside, the user can adjust the pronunciation, accent, intonation, breath, etc. independently adjusted by the user. It is possible to obtain a desired synthesized sound with the following, etc.

Moreover, since the present invention can be realized by simply adding a mode setting section and a control section, the configuration of the device is simplified, and the effects of reducing the circuit size and cost can be expected.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of a text-to-speech conversion device according to an embodiment of the present invention, and FIG. 2 is a block diagram of a conventional text-to-speech conversion device. 1...Host computer, 20...
Text/speech conversion device, 21...interface unit, 22...word dictionary memory, 30...
... text analysis section, 40 ... mode setting section, 50 ... control section, 51 ... mode judgment means, 52 ... input/output selection means , 60...
...Speech synthesis section, 61...Speaker.

Claims (1)

[Scope of Claims] A text analysis unit that analyzes kanji-kana mixed text data and generates a phonological and prosodic symbol string for the text data;
A text-to-speech conversion device comprising a speech synthesis unit that outputs synthesized speech corresponding to the text data based on the phoneme and prosodic symbol strings, a text-to-speech conversion mode in which the text data is input to the text analysis unit. , a phonological and metrical symbol string output mode in which the phonological and metrical symbol strings generated by the text analysis section are outputted to the outside in the form of character codes, and a phonological and metrical symbol string from the outside is directly input to the speech synthesis section. a mode setting unit that externally sets one of the phoneme and prosodic symbol string synthesis modes; and a mode setting unit that determines the mode state set by the mode setting unit and converts the text data and the phoneme and prosodic symbol strings based on the determination result. 1. A text-to-speech conversion device comprising: a control section for controlling input and output of the text.