JPH0954577A - Voice generator - Google Patents

Abstract

PROBLEM TO BE SOLVED: To generate the voices corresponding to assigned words without previously requiring to store the sound source codes corresponding to all the words with a voice generator for language. SOLUTION: The pronunciation signal corresponding to the retrieved English word is read out when a Japanese language is inputted by the character key 12a of, for example, a key input section 12 and the English word which is the Japanese equivalent to this English language is retrieved from a dictionary memory 13a in a ROM 13 according to the operation of a 'translation' key 12d. This pronunciation signal is converted to the sound source cord for generating voices. The code is stored in a sound source code register 15a in the RAM 14. The parameters (length and size) of the accent corresponding to the pronunciation signal added with an accent are read out of an accent parameter memory 13b when the pronunciation signal corresponding to the retrieved English word is judged to be added with the accent. The voices are generated in a voice synthesis section 18 according to the sound source code corresponding to the pronunciation signal of the retrieved word and the parameters of the accent and are outputted from a speaker 19.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a speech processing apparatus such as an electronic dictionary apparatus for generating a speech of a language.

[0002]

2. Description of the Related Art Conventionally, for example, when one wants to know English corresponding to Japanese or Japanese corresponding to English, a Japanese-English dictionary or an English-Japanese dictionary has been used, but recently, the dictionary data is stored in a ROM or the like. An electronic dictionary device that stores in a storage medium such as a semiconductor memory or a compact disk and is accessed by using an electronic device, for example, when a Japanese key is input, the corresponding English language is searched and displayed and output on the display unit. It has been put to practical use.

In such an electronic dictionary device, a device having a function of generating a voice in a designated language has been put into practical use. That is, in the voice generation function of this electronic dictionary device, for example, when Japanese is input and the corresponding English is searched and displayed and output, the pronunciation of the English is also output as voice. .

In the conventional voice generation function, sound source codes corresponding to all the words included in the dictionary data are stored in advance as a means for generating a voice in a specified language, and a certain word is stored. When is specified, the sound source (speech synthesis circuit) is driven according to the sound source code corresponding to the specified word, and the sound of the word is output.

[0005]

As described above, conventionally,
For example, in order to generate the voice of an English word, the code data for driving the sound source must be stored in advance corresponding to each of all the words included in the dictionary data. There is a problem that a large storage area is required for generation.

The present invention has been made in view of the above problems, and it becomes possible to generate a voice corresponding to a designated word without pre-storing sound source codes corresponding to all the words. An object is to provide a voice generating device.

[0007]

That is, a speech generating apparatus according to claim 1 of the present invention comprises a language data storage means for storing a language and its phonetic symbols, and a phonetic symbol stored in this language data storage means. Accent parameter storage means for storing in advance the accent parameter corresponding to, the designation means for arbitrarily designating a word from the language data storage means, and the phonetic symbol corresponding to the word designated by the designating means for the language data storage. Phonetic symbol reading means for reading from the means, accent judging means for judging whether or not each of the phonetic symbols read by the phonetic symbol reading means is accented, and the phonetic symbol reading means by the accent judging means. If it is determined that the phonetic symbol read by is accented, Accent parameter reading means for reading from the accent parameter storage means the parameter of the accent corresponding to the accented phonetic symbol, and the phonetic symbol obtained by the phonetic symbol reading means and the accent parameter reading means. A voice generating means for generating a voice corresponding to the word designated by the designating means based on the accent parameter.

That is, in the speech generating apparatus according to the first aspect, when a word is arbitrarily designated from the language data storage means storing the language and the pronunciation symbol, the pronunciation symbol corresponding to the designated word is read out and the pronunciation is pronounced. If it is determined that the symbol is accented, the accent parameter corresponding to the accented phonetic symbol is
The accent parameter corresponding to each phonetic symbol stored in the language data storage means is read in advance from the accent parameter storage means, and based on the read phonetic symbol and the read accent parameter. Since the voice corresponding to the specified word is generated, the phonetic symbols originally existing with the language in the language data storage means should be the basis of the sound source, and only the accent parameters corresponding to each phonetic symbol should be stored as table data. It is sufficient to add intonation to the voice corresponding to the specified language according to the accent parameter.

According to a second aspect of the present invention, there is provided a voice generating device, which stores in advance dictionary data in which dictionary data including a plurality of words and phonetic symbols thereof are stored, and accent parameters corresponding to the phonetic symbols in advance. The accent parameter storage means, the pronunciation symbol reading means for reading out the pronunciation symbol corresponding to the designated word when the word stored in the dictionary memory is designated, and the pronunciation symbol reading means for reading the pronunciation symbol. Accent judging means for judging whether or not each of the phonetic symbols is accented, and when the accent judging means judges that the phonetic symbols read by the phonetic symbol reading means are accented. Is the parameter of the accent corresponding to the accented phonetic symbol. The accent parameter reading means for reading from the data storage means, and the voice generating means for generating a voice based on the pronunciation symbol obtained by the pronunciation symbol reading means and the accent parameter read by the accent parameter reading means. It is characterized by

That is, in the voice generating device according to the second aspect, when a word stored in the dictionary memory is designated,
The phonetic symbol corresponding to the designated word is read.
Then, when it is determined that the phonetic symbol corresponding to the designated word is accented, the accent parameter corresponding to the accented phonetic symbol is the accent parameter corresponding to each phonetic symbol. Parameters are read out from the accent parameter storage means in advance, and a voice is generated based on the phonetic symbol of the specified word and the accent parameter read out from the accent parameter storage means. It suffices that the symbols be the basis of the sound source, and only the parameters of the accents corresponding to the individual phonetic symbols are stored as table data, and the generated speech will be added with intonation by the parameters of the accents.

A voice generating apparatus according to claim 3 of the present invention is the voice generating apparatus according to claim 1 or 2, wherein the parameter is the length of the voice generated by the voice generating means. And a parameter for specifying the size.

That is, in the voice generating device according to the third aspect, in the voice generating device according to the first or second aspect, the parameters thereof are the length and volume of the voice generated by the voice generating means. For example, in claim 1, the phonetic symbol originally existing with the language in the language data storage means is used as the basis of the sound source, and the length of the voice of the accent corresponding to each phonetic symbol is defined. It suffices to store only the parameter specifying the and the loudness as table data, and the voice corresponding to the designated language is added with the intonation consisting of the length and the loudness of the voice depending on the parameter of the accent.

[0013]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing a configuration of an electronic circuit of an electronic dictionary device according to an embodiment of a voice generating device of the present invention.

This electronic dictionary device has a CPU 11. The CPU 11 activates a system program stored in advance in the ROM 13 in response to a key operation signal from the key input unit 12, and controls the operation of each part of the circuit in accordance with this system program. RAM 14 in addition to the unit 12 and the ROM 13
And the liquid crystal display unit 17 is connected via the display drive circuit 16.

Further, the CPU 11 includes a voice synthesizing unit 1
A speaker 19 is connected via 8. Key input unit 12
Is a letter key 12a, "Japanese-English" key 12b, "English-Japanese"
A key 12c, a "translation" key 12d, a "pronunciation" key 12e and the like are provided.

The character key 12a includes a kana key for inputting Japanese and English, a kana / kanji conversion key, and an English character key. The "Japanese-English" key 12b is operated when setting the Japanese-English conversion mode from Japanese to English.

The "English-Japanese" key 12c is operated when setting the English-Japanese conversion mode from English to Japanese. The "translation" key 12d is operated when instructing the execution of the English-Japanese or Japanese-English conversion.

The "pronunciation" key 12e is operated when outputting the pronunciation of the input English word or the translated English word as a voice. In the ROM 13, various subprograms such as a Japanese input processing program, an English input processing program, a translated word search processing program, and a pronunciation data creation processing program are stored in advance, together with the system program that controls the operation of the entire electronic dictionary device.

Further, the ROM 13 is provided with a dictionary memory 13a in which dictionary data in which a large number of Japanese words, English words, and their phonetic symbols are associated with each other is stored in advance, and accents corresponding to each phonetic symbol are provided. Parameter memory 13b in which the parameters of
Is provided.

FIG. 2 is a diagram showing a storage state of accent parameters for each phonetic symbol in the accent parameter memory 13b of the electronic dictionary device. That is, the accent parameter memory 13b stores in advance the intonation parameters including the pronunciation length and size corresponding to each phonetic symbol when accents are added to all existing phonetic symbols.

The RAM 14 is provided with a display register 14a, an input character storage section 14b, an address register 14c, and a sound generation register 15 (sound source code register 15a, parameter register 15b).

The display register 14a includes a liquid crystal display unit 17
The display data to be displayed at is written as bitmap format data. In the input character storage unit 14b, Japanese or English input characters input by the character keys 12a of the key input unit 12 are written.

The address register 14c has a Japanese-English conversion mode or a "translation" key 12d in the English-Japanese conversion mode.
The memory address of the English word or the Japanese word retrieved from the dictionary memory 13a in the ROM 13 in accordance with the operation of is written.

The sound source code register 15a of the pronunciation generation register 15 corresponds to the pronunciation symbols of the English words searched for as translations, which are stored in the memory addresses of the dictionary memory 13a written in the address register 14c. It is converted into a sound source code and stored.

The parameter register 15b of the tone generation register 15 stores accent parameters corresponding to each tone code stored in the tone code register 15a.

That is, in the parameter register 15b, the accents corresponding to the phonetic symbols of the English words searched as the translated words, which are stored in the memory addresses of the dictionary memory 13a written in the address register 14c, are stored. Parameters are stored, in this case
Among the phonetic symbols of the searched English words, the parameters (length, size) of phonetic symbols with accents are stored in the ROM.
All the variation amounts of the parameters of the phonetic symbols read and written from the accent parameter memory 13b (see FIG. 2) in 13 and to which other accents are not added are "± 0".

The liquid crystal display unit 17 displays Japanese, English, or their translations entered by the key input unit 12. The voice synthesizing unit 18 produces a voice signal in English according to the tone generator code stored in the tone generator code register 15a in the pronunciation generation register 15 in the RAM 14 and the accent parameter stored in the parameter register 15b corresponding to each tone generator code. The voice signal generated by the voice synthesizer 18 is output as voice from the speaker 19.

Next, the operation of the electronic dictionary device having the above configuration will be described. FIG. 3 is a flowchart showing the overall processing of the electronic dictionary device. FIG. 4 is a flowchart showing a pronunciation data creation process of the electronic dictionary device.

For example, when wanting to know English corresponding to Japanese, operate the "Japanese-English" key 12b of the key input unit 12 to set the Japanese-English conversion mode, and operate the character key 12a to input the desired Japanese. Then, it is determined that the mode is the Japanese-English conversion mode, the input Japanese character is displayed on the liquid crystal display unit 17, and the input character storage unit 14b in the RAM 14 is displayed.
Is written in (step S1 → S2 → S4).

On the other hand, for example, when wanting to know Japanese corresponding to English, the "English-Japanese" key 12b of the key input unit 12 is operated to set the English-Japanese conversion mode, and the character key 12a is operated to display the desired English. When input, it is determined that the mode is the English-Japanese conversion mode, the input English characters are displayed on the liquid crystal display unit 17, and the input character storage unit 14 in the RAM 14 is displayed.
It is written in b (steps S1 → S2 → S3).

Then, for example, when the Japanese-English conversion mode is set, "Ai" is input in Japanese and the "Translation" key 12d is entered.
When is operated, the English word corresponding to the input Japanese "love" is searched from the dictionary memory 13a in the ROM 13 (steps S5 → S6).

Here, in response to the input Japanese "love", "LOV" is stored in the dictionary memory 13a in the ROM 13.
When "E" is searched, the memory address is determined to be RAM14.
Are stored in the internal address register 14c (step S
7 → S8).

Then, the English word "LOVE" corresponding to the Japanese word "love" retrieved in the dictionary memory 13a is displayed on the liquid crystal display unit 17 (step S9). Thus
The English word “LOVE” converted from Japanese to English is displayed on the liquid crystal display unit 1.
When the user wants to hear the pronunciation of the English word as a voice while it is displayed on 7, the "pronunciation" key 12e of the key input unit 12 is displayed.
Is operated, the address register 14c in the RAM 14 is operated.
It is determined whether or not the memory address of the English word to be pronounced is stored in (step S10 → S11).

In step S11, "YES",
That is, since it is determined that the add register 14c in the RAM 14 stores the memory address in the dictionary memory 13a of the English word "LOVE" retrieved in association with the Japanese-English conversion process, the pronunciation data creation in FIG. The process shifts to (step S11 → SA).

In the pronunciation data creation process in FIG. 4, first, the dictionary memory 13a in the ROM 13 is based on the memory address stored in the address register 14c.
The phonetic symbol "l" of the English word "LOVE" stored in
The first phonetic symbol “l” of Λ′V ”(the accent symbol“ ′ ”is assigned corresponding to“ Λ ”) is read and converted into a sound source code for voice generation. The tone generator code register 15a of the pronunciation generation register 15
(Steps A1, A2, A3).

Here, it is judged whether or not the first phonetic symbol "l" read from the dictionary memory 13a is accented (step A4). In this case, it is determined that the first phonetic symbol "l" is not accented, and the process returns to the phonetic symbol reading process from step A1 (step A4 → A7 → A1).

Then, the dictionary memory 1 in the ROM 13
Of the phonetic symbols "lΛ'V" of the English word "LOVE" stored in 3a, the second phonetic symbol "Λ '" is read, converted into a sound source code for voice generation, and stored in the sound generation register 15. It is stored in the sound source code register 15a (steps A1, A2, A3).

In this case, the second phonetic symbol "Λ '"
Since it is determined that the accent parameter has an accent, the accent parameter corresponding to the accented phonetic symbol “Λ ′” has a length of “−” from the accent parameter memory 13b (see FIG. 2) in the ROM 13. 5 "and size" +5 "are read and stored in the corresponding parameter register 15b (steps A4 → A5, A6).

In this way, when the conversion process to the sound source code corresponding to the second phonetic symbol "Λ '" and the reading setting process of the accent parameter thereof are performed, the phonetic symbol reading process from step A1 is further performed. Return (step A7 → A1).

Then, the dictionary memory 1 in the ROM 13
Of the phonetic symbols "lΛ'V" of the English word "LOVE" stored in 3a, the third phonetic symbol "V" is read and converted into a sound source code for sound generation, and the sound source of the sound generation register 15 is read. It is stored in the code register 15a (steps A1, A2, A3).

In this case, it is determined that the third phonetic symbol "V" is not accented, and that it is the last phonetic symbol corresponding to the retrieved English word "LOVE". Therefore, the conversion process to the sound source code corresponding to each of the phonetic symbols and the accent parameter setting process are ended (steps A4 → A7).

That is, the sound source code register 15a of the sound generation register 15 is searched and the searched English word "LO" is read.
The sound source code corresponding to the phonetic symbol "lΛ'V" of "VE" is stored, and the parameter register 15b stores an accent parameter indicating the intonation (length, loudness) of the voice corresponding to the addition position of the accent symbol. Accordingly, in the voice generation unit 18, the voice corresponding to the voice source code stored in the voice source code register 15a is converted into a voice (long voice) according to the accent parameter stored in the parameter register 15b. Size
And is output from the speaker 19 (step A7).
→ A8).

In this case, it is not necessary to previously store the sound source codes individually corresponding to all the English words stored in the dictionary memory 13a, and the phonetic symbols of the searched English words are converted into the sound source codes, Only for the sound source code in which the phonetic symbol is accented, the accent parameter of the phonetic symbol is read from the accent parameter memory 13b and added, and the intonation at the time of sound generation by the sound source code is added, so that all the pronunciations are pronounced. Only the accent parameter corresponding to the symbol is accent parameter memory 13b.
If stored as, it is possible to easily generate pronunciations of all the English words stored in the dictionary memory 13a.

On the other hand, in FIG. 3, the key input unit 12
By operating the character key 12a of, for example, while the desired English word “LOVE” is being input, the “translation” key 12
If the "pronunciation" key 12e is operated without operating d, RA
The dictionary memory 13 is stored in the address register 14b in M14.
It is determined that the memory address indicating the word in a is not stored (step S10 → S11).

Then, the input English word "LOV
"E" is retrieved from the dictionary memory 13a, and its memory address is written in the address register 14b (steps S11 → S12, S13 → S14).

Thus, the address register 14b
When the memory address indicating the English word "LOVE" stored in the dictionary memory 13a in the ROM 13 is written in, the process shifts to the pronunciation data creation process in FIG. Similar to the pronunciation data creation process corresponding to a word, the reading process of each phonetic symbol and the conversion process to the sound source code, the accent parameter read setting process, the voice generation process according to the sound source code and the accent parameter are executed, and The pronunciation of the English word "LOVE" that is keyed in is output as voice from the speaker 19 (step SA).

On the other hand, the "translation" key 1 of the key input unit 12
The search process of the dictionary memory 13a according to the operation of 2d (steps S5 to S6) or the search process of the dictionary memory 13a of the keyed-in English word itself (step S10).
If the corresponding English word is not found in S12), an error message is displayed on the liquid crystal display unit 17 (step S7 → S15 or step S13 → S1).
5).

Therefore, according to the electronic dictionary device having the above-mentioned configuration, for example, Japanese is input by the character keys 12a of the key input unit 12, and the English words which are the translated words of the Japanese are accompanied by the operation of the "translation" key 12d. Is the dictionary memory 1 in the ROM 13
3a, the phonetic symbol corresponding to the searched English word is read out, converted into a sound source code for voice generation, stored in the sound source code register 15a in the RAM 14, and corresponds to the searched English word. When it is determined that the accented phonetic symbol is accented, the accent parameter (length, size) corresponding to the accented phonetic symbol is read from the accent parameter memory 13b, and the search is performed. Since a voice is generated by the voice synthesizer 18 according to the sound source code corresponding to the phonetic symbol of the word and the accent parameter and output from the speaker 19, it corresponds to all the English words stored in the dictionary memory 13a individually. It is not necessary to store the sound source code in advance, but the phonetic symbols of the searched English words are converted into sound source codes and the Only for sound source codes whose symbols are accented, accent parameters of phonetic symbols can be read from the accent parameter memory 13b and added, and intonation at the time of sound generation by the sound source codes can be added. If only the accent parameter corresponding to is stored in the accent parameter memory 13b, pronunciations of all the English words stored in the dictionary memory 13a can be easily generated.

[0049]

As described above, according to the voice generating device of the first aspect of the present invention, when a word is arbitrarily designated from the language data storage means that stores the language and the phonetic symbol, it corresponds to the designated word. When the phonetic symbol is read out and it is determined that the phonetic symbol is accented, the accent parameter corresponding to the accented phonetic symbol is stored in the language data storage means. A voice corresponding to the specified word is read from the accent parameter storage means that stores in advance accent parameters corresponding to the phonetic symbols, and based on the read phonetic symbols and the read accent parameters. Since it is generated, the phonetic symbols originally existing with the language in the language data storage means are the basis of the sound source,
Only the parameter of the accent corresponding to each phonetic symbol needs to be stored as table data, and the voice corresponding to the designated language is added with the intonation according to the parameter of the accent.

According to the second aspect of the present invention, when the word stored in the dictionary memory is designated, the phonetic symbol corresponding to the designated word is read. Then, when it is determined that the phonetic symbol corresponding to the designated word is accented, the accent parameter corresponding to the accented phonetic symbol is the accent parameter corresponding to each phonetic symbol. Parameters are read out from the accent parameter storage means in advance, and a voice is generated based on the phonetic symbol of the specified word and the accent parameter read out from the accent parameter storage means. It is only necessary to store the symbols as the basis of the sound source and only the parameters of the accents corresponding to the individual phonetic symbols are stored as table data, and the tones generated are added with the intonation by the parameters of the accents.

According to a third aspect of the present invention, there is provided the voice generating device according to the first or second aspect, wherein the parameter is the voice generated by the voice generating means. Since the parameters are used to specify the length and the size, for example, in claim 1, the phonetic symbols originally existing in the language data storage means together with the language are used as the basis of the sound source, and the accents corresponding to the individual phonetic symbols are provided. Only the parameters that specify the length and volume of the voice need to be stored as table data, so that the voice corresponding to the designated language is added with the intonation consisting of the length and the volume of the voice depending on the accent parameter. Become.
Therefore, according to the present invention, it is possible to generate a voice corresponding to a designated word without pre-storing sound source codes corresponding to all the words.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of an electronic circuit of an electronic dictionary device according to an embodiment of a voice generating device of the present invention.

FIG. 2 is a diagram showing a storage state of accent parameters for each phonetic symbol in an accent parameter memory of the electronic dictionary device.

FIG. 3 is a flowchart showing the overall processing of the electronic dictionary device.

FIG. 4 is a flowchart showing a pronunciation data creation process of the electronic dictionary device.

[Explanation of symbols]

 11 ... CPU, 12 ... Key input part, 12a ... Character key, 12b ... "Japanese-English" key, 12c ... "English-Japanese" key, 12d ... "Translation" key 12e ... "Pronunciation" key, 13 ... ROM, 13a ... Dictionary memory , 13b ... Accent parameter memory, 14 ... RAM, 14a ... Display register, 14b ... Input character storage section, 14c ... Address register, 15 ... Voice creation register, 15a ... Sound source code register, 15b ... Parameter register, 16 ... Display drive circuit , 17 ... Liquid crystal display section, 18 ... Voice synthesis section, 19 ... Speaker.

Claims (3)

[Claims] 1. Language data storage means for storing a language and its phonetic symbols, accent parameter storage means for storing in advance accent parameters corresponding to each phonetic symbol stored in this language data storage means, said language data Designating means for arbitrarily designating a word from the storage means, phonetic symbol reading means for reading a phonetic symbol corresponding to the word designated by the designating means from the language data storage means, and a phonetic symbol reading means for reading the phonetic symbol. Accent judging means for judging whether or not each of the phonetic symbols is accented, and when the accent judging means judges that the phonetic symbols read by the phonetic symbol reading means are accented. Is the accent parameter corresponding to the accented phonetic symbol. Is read from the accent parameter storage means, and the word designated by the designating means is based on the phonetic symbol obtained by the phonetic symbol reading means and the accent parameter read by the accent parameter reading means. A voice generation device comprising: a voice generation unit that generates a corresponding voice. 2. A dictionary memory in which dictionary data consisting of a plurality of words and phonetic symbols thereof are stored in advance, an accent parameter storage unit in which accent parameters corresponding to each phonetic symbol are stored in advance, and the dictionary memory stores the data. And a phonetic symbol reading means for reading out a phonetic symbol corresponding to the specified word, and whether or not each of the phonetic symbols read by the phonetic symbol reading means is accented. And an accent judgment means for judging whether the phonetic symbol read by the phonetic symbol reading means is accented or not, it corresponds to the accented phonetic symbol. An accent parameter for reading accent parameters from the accent parameter storage means. Sound generating device comprising a read unit, by comprising a sound generating means for generating a voice based on the parameter of the read-out accent by the resulting phonetic symbols by said pronunciation symbol reading means the accent parameter reading means. 3. The voice generating device according to claim 1, wherein the parameter is a parameter for designating a length and a volume of a voice generated by the voice generating means.