JPH09134195A - Speech synthesizer - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain a speech synthesizer capable of decreasing a processing amt. and reducing a cost by forming the character strings of reading with accent and phoneme symbols from the sentence examples selected from a sentence example section and the character code strings of the proper nouns outputted from a KANA (Japanese syllabary) reading conversion section. SOLUTION: The sentences of guide information are outputted dividedly to the fixed plural guide commands and the character code strings of the variable proper nouns and figures to the speech synthesizer from a navigation ECU 11. The character table section 2 in the speech synthesizer stores the plural fixed sentence examples in response to the guide commands from the navigation ECU 11 and outputs the same. The KANA (Japanese syllabary) reading conversion section 3 forms variable reading sentence examples in accordance with the proper nouns, figures and code strings. A section 4 for forming the reading characteristic strings with the accent and phonemes imparts the accents and phonemes to the variable reading sentence examples from the reading conversion section 3 and synthesizes these sentence examples with the fixed sentence examples from the sentence example table section 2. As a result, the speech guide including the arbitrary place-names and figures is embodied with a simple constitution.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a text-to-speech synthesizer, and more particularly to a voice-synthesizer capable of providing voice information by making only proper nouns and numbers variable by using limited sentence examples.

[0002]

2. Description of the Related Art In a conventional text-to-speech synthesizer, a language processing unit is provided to perform grammatical interpretation of input text, understand meaning, etc., and add accents and intonations.

[0003]

However, in the conventional text-to-speech synthesizer, since the algorithm of the language processing unit is complicated and the processing amount is large, it is adopted in a system or the like operating on a personal computer. There is a problem that it is practically difficult to mount the system on an inexpensive system.

Therefore, in view of the above problems, it is an object of the present invention to provide a speech synthesizer which can reduce the amount of processing and can be manufactured at low cost.

[0005]

In order to solve the above problems, the present invention provides a speech synthesizer having the following configuration. That is, in a voice synthesizing device for voice synthesizing character guidance information from a navigation installed in a vehicle, variable information having the character guidance information as a proper noun and a numerical code string and fixed sentence example (number) information as a guidance command are provided. A navigation ECU which outputs the fixed information as a sentence example, and a sentence example table section which outputs a sentence example corresponding to the guidance command and a character code string of a proper noun of the variable information. A kana-yomi conversion unit for converting into a kana-yomi character string, a sentence example selected from the sentence example table part, and a character string of a proper noun output from the kana-yomi changing unit to generate a reading character string with accents and prosodic symbols. And a reading character string generating section with accent prosody. With this device, voice guidance including an arbitrary place name can be realized with a simple configuration.

The Kana-yomi conversion unit further converts a numerical character code string into a reading character string, and the accent-prosodic reading character string generation unit further converts the sentence example selected from the sentence example table unit and the Kana-Yomi conversion unit. A reading character string with accent and prosody symbols may be generated from a character code string of numbers output from the department. With this device, voice guidance including an arbitrary number can be realized with a simple configuration.

The accented prosodic reading character string generation unit depends on the mora length of the proper noun output from the Kana-yomi conversion unit, and the pose length between the proper noun and the sentence example selected from the sentence example table unit. May be changed to perform composition. By this means, it is possible to provide a synthetic sound that is easy to hear. The accented prosodic reading character string generation unit selects a sentence example selected from the sentence example table unit that follows the proper noun output from the Kana-yomi conversion unit, with the accent side of the proper noun such as “near”. The accent type of a proper noun may be set to O type when it starts with a suffix that has the effect of making it O type. By this means, a natural synthetic sound can be generated.

In the accented prosodic reading character string generation unit, the sentence example selected from the sentence example table unit following the proper noun output from the Kana-Yomi conversion unit is
When it does not start with a suffix such as “near” that has the effect of making the accent side of a proper noun into an O type, the accent nucleus may be placed at the third mora from the end of the proper noun. By this means, a natural synthetic sound can be generated.

When the third mora from the ending of the proper noun is either "long sound" or "prompt", the accent nucleus may be shifted and the accent nucleus may be placed at the fourth mora from the ending. By this means, a natural synthetic sound can be generated. The accented prosodic reading character string generation unit may insert a pause length at a predetermined break in the proper noun when the mora length of the proper noun output from the Kana-yomi conversion unit is large.

The accented prosodic reading character string generation unit depends on the mora length of the code string of the number output from the Kana reading conversion unit, and the pause length between the number and the sentence example selected from the sentence example table unit. May be changed to perform composition. By this means, a natural synthetic sound can be generated. The accented prosodic reading character string generation unit is a sentence example selected from the sentence example table unit following the numerical code string output from the Kana reading conversion unit,
When starting with a suffix that acts to make the accent side of the numbers O-type, such as “degree” and “first”, the accent type of the number string may be O-type. By this means, a natural synthetic sound can be generated.

In the accented prosodic reading character string generation unit, the sentence examples selected from the sentence example table unit subsequent to the numerical code string output from the Kana reading conversion unit are "degree" and "first". When the suffix does not start with a suffix that acts to make the accent side of the number O-shaped, the accent nucleus may be placed at the third mora from the end of the code sequence of the number. By this means, a natural synthetic sound can be generated.

When the third mora from the end of the numerical code sequence is either "long sound" or "consonant", the position of the accent nucleus may be shifted to place the accent nucleus at the fourth mora from the end. . By this means, a natural synthetic sound can be generated. When the code string of the numerical value is time, "0
3 from the ending of "00" at the time indicated by "0:00"
The accent nucleus may be placed in the mora, and the accent nucleus position may be shifted to the fourth mora from the end when the third mora from the end is either "long sound" or "prompt". By this means, a natural synthetic sound can be generated.

When the numerical code string is the time, "00"
3 from the ending of "00 hour" at the time indicated by "hour △ minute"
The accent nucleus may be placed on the "ji" of the mora eyes. By this means, a natural synthetic sound can be generated. When the code sequence of the numeral includes a decimal point, an accent nucleus is placed in the third moor from the ending of “00.” indicated by “00.ΔΔ”, and the third mora from the ending is “long sound” and “consonant”. In either case, the accent nucleus position may be shifted to the fourth mora from the end of the word. By this means, a natural synthetic sound can be generated.

When the proper noun / numeral code string output from the kana-yomi conversion unit includes a "ga" line sound, it may be converted to a nasalized "ga" line sound. By this means, a natural synthetic sound can be generated. When the code string of proper nouns and numbers output from the kana-yomi conversion unit includes a continuation of the same vowel, the continuation of the vowel may be extended. By this means, a natural synthetic sound can be generated.

When the proper noun and numeric code string output from the Kana-yomi conversion unit includes a vowel sequence of "EI", the vowel sequence is made to be long like the same vowel sequence of "EE". May be. When the proper noun / numeric code string output from the kana-yomi conversion unit includes a vowel continuation of "OU", the vowel continuation may be extended as in the case of the vowel continuation of "OO". By this means, a natural synthetic sound can be generated. By performing the simple accent processing in this manner, it is possible to provide a synthetic sound with intonation and accent added without receiving accent or prosody information from the outside.

[0016]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a diagram showing a speech synthesizer according to the present invention. The speech synthesizer according to the present invention is used for a navigation system mounted on a vehicle, and the navigation system provides various information by changing only proper nouns, numbers, etc. using limited sentence examples. To do. As shown in this figure, the navigation ECU
From (Electronic Control Unit) 11 to the voice synthesizer, a sentence of guidance information is provided with a plurality of fixed guidance commands, a variable proper noun, and a numeric character code string (JIS, S-JIS).
Etc.) and are output separately. The voice synthesizer stores a sentence example table 2 for storing and outputting a plurality of fixed sentence examples in response to a guidance command from a navigation ECU (Electronic Control Unit) 11, and a navigation ECU (Electronic Electronic).
Kana Yomi conversion unit 3 that forms a variable Yomi sentence example based on proper nouns, numbers, and code strings (JIS, S-JIS, etc.) from Control Unit 11 and accents and prosody on the variable Yomi sentence example of Yomi conversion unit 3. An accented prosodic reading character string generation unit 4 that adds and synthesizes a fixed sentence example from the sentence example table 2
, An acoustic processing unit 5 that acoustically processes the synthesized sentence example from the accented prosodic reading character string generation unit 4, a phoneme dictionary unit 6 for providing phonemes to the acoustic processing unit 5, and an accent rule to the acoustic processing unit 5. An accent rule unit 7 to be provided, a prosody rule unit 8 to provide a prosody rule to the sound processing unit 5, and a P to convert processing data from the sound processing unit 5 into PCM data.
A CM data converter 9 is provided. In this way, voice guidance including arbitrary place names and numbers can be realized with a simple configuration.

Here, the sentence example table section 2 of the speech synthesizer is used.
Uses the sentence number as a parameter for the guidance command input from the navigation ECU 1, as shown in the table below.
It has a plurality of fixed reading character examples in which a reading character string is provided with accent and prosody symbols.

[0018]

[Table 1]

"/" Indicates the mora boundary position,
"A (circle)" indicates the accent nucleus position, "," indicates the phrase end position, and "." Indicates the end of the sentence. A pose length is provided at each of these positions, and "/", "a"
(Circle) ”,“, ”and“. ”Increase in that order. further,
The kana-reading conversion section 3 responds to the variable kana proper nouns and numeric code strings sent from the navigation ECU 1.
Convert it to a reading string as shown in the table below.

[0020]

[Table 2]

Further, the accented prosodic reading character string generation unit 4 stores the fixed sentence example selected from the character string table unit 2 and the variable reading character string output from the kana conversion unit 3.
Synthesize as shown in the table below.

[0022]

[Table 3]

In this case, the variable reading character string from the kana conversion unit 3 is in the range of 1 character to 20 or more mora. When synthesizing the neighborhood "The guidance point is near XX", the accented prosodic reading character string generation unit 4 uses "XX" and "near" for the small mora length of the proper noun corresponding to "XX". A large pause length may be inserted between "". In particular,
The shorter the word, such as "tsu (1 mora)", the larger the pose length to make it easier to hear.

Regarding the time, a large pause length may be inserted between "3" and "minutes" for "there is XX minutes". Furthermore, regarding the distance,
A large pause length may be inserted between "1" and "km" for "m." Next, the variable reading character string of the proper noun sent from the kana-yomi conversion unit 3 has no accent information, and when the mora length of this character string is large, it is difficult to hear without accent. Therefore, in general, the accent nucleus is often located at the 3rd mora from the end of the word, so the accent may be placed here. For example,
In the case of "Iidabashi (Iidabashi)", it is possible to generate a natural synthesized sound by placing an accent nucleus at "da" which is the third mora from the ending. Further, the accent type of the numerical string such as “degree” or “preceding” is set to “O type” for the sentence example (character string with accent / prosodic symbol held in advance in the synthesizing device) subsequent to the numerical code string. If it does not start with a suffix having an action, an accent nucleus may be placed at the third mora from the end of the proper noun. This is because it becomes possible to obtain a natural synthesized sound as well.

However, when a suffix such as "near" in the fixed sentence example is added, the position of the accent nucleus is placed in "near", and "Iidabashi" becomes a flat (O type) accent type. An example of a sentence following the numeric code string (character string with accent / prosodic symbol stored in the synthesizer in advance)
Indicates that the accent type of a number string such as "degree" or "destination" is "O"
When starting with a suffix that has the effect of making "type", a natural synthetic sound can be generated by changing the accent type of the number string to "O type and flat type".

[Hanshin Kosoku (Hanshin Expressway)]
As described above, when the third mora from the end is “long sound” and the “consonant”, the position of the accent nucleus is shifted by one in the beginning direction and placed at the fourth mora from the end. When the third mora from the end of the numerical code string is “long sound” or “consonant”, the position of the accent nucleus may be shifted to place the accent nucleus at the fourth mora from the end. This is because it is possible to obtain a natural synthesized sound as well. The above table is summarized below.

[0027]

[Table 4]

Next, the proper noun of the variable reading character string from the kana conversion unit 3 has a long mora length such as "National Highway No. 2" and "Ginza 4-chome", and the meaning in the proper noun has a break. For words, the pause length is inserted at this break as shown in the table below, and a natural synthesized voice is generated.

[0029]

[Table 5]

Next, in the accented prosodic reading character string generation unit 4, when the following numerical code string is input from the Kana-Yomi conversion 3, the accent is not given at the third mora from the end but at a predetermined place. Put a nucleus. About the time
In the case of “10:10”, an accent nucleus may be placed at the third mora from the end of “Juji”, which represents time, as in “Ju'uji Jupun”.

Regarding time, if "1 hour 15 minutes",
An accent nucleus may be placed at the third mora from the end of "Ijijikan", which represents time, such as "Ijijikanjugofun". In the case of "4.5" for a small number, an accent nucleus may be placed at the beginning of the word, as in "Yontengo".

Next, in the accented prosodic reading character string 4, when a "ga" line sound is input from the kana-yomi conversion 3, the "ga" line sound has a very high probability of becoming nasal. All the "ga" line sounds of the received character string are processed as nasal noise. Next, in the accented prosodic reading character string 4, if the character string sent from the Kana-yomi conversion 3 contains a sequence of the same vowels, it is lengthened. For example, it may be possible to lengthen the sound such as “Osaka” → “Osaka”.

Next, in the accented prosodic reading character string 4, "E" is added to the character string sent from the Kana reading conversion 3.
When the continuous vowels of "I" and "OU" are included, they are lengthened. For example, “Meiji” → “Meiji” “DOURO” → “Doro” In this way, by performing simple accent processing, accent and prosody information is not received from outside. Together, it is possible to provide a synthetic sound with intonation and accent.

[Brief description of the drawings]

FIG. 1 is a diagram showing a speech synthesizer according to the present invention.

[Explanation of symbols]

 1 ... Navigation ECU 2 ... Sentence table section 3 ... Kana-reading conversion section 4 ... Accent prosodic reading character string

Claims (18)

[Claims] 1. A voice synthesizing device for synthesizing voice guidance information from a navigation mounted on a vehicle, wherein the text guidance information is variable character string information including proper nouns and numeric code strings, and fixed sentence examples as guidance commands. A navigation ECU (1) that outputs (number) information separately, and a sentence example table unit (2) that outputs a sentence example corresponding to the guidance command by storing a plurality of the fixed information as a sentence example into a table and storing the table. A kana-yomi conversion unit (3) for converting a character code string of the proper noun of the variable information into a kana-yomi character string, a sentence example selected from the sentence example table unit (2), and an output from the kana-reading changing unit (3). And a accented prosodic reading character string generation unit (4) for generating a reading character string with accent and prosody symbols from a character code string of a proper noun. Apparatus. 2. The kana-yomi conversion unit (3) further converts a numeric character code string into a reading character string, and the accented prosodic reading character string generation unit (4) further includes the sentence example table unit (2). 2. The speech according to claim 1, wherein a reading character string with accents and prosodic symbols is generated from a sentence example selected from) and a character code string of numbers output from the kana reading changing unit (3). Synthesizer. 3. The accented prosodic reading character string generation unit (4) depends on the proper noun and the sentence example table unit (2) depending on the mora length of the proper noun output from the Kana reading conversion unit (3). 2. The speech synthesis apparatus according to claim 1, wherein the speech synthesis is performed by changing the pause length between the sentence example selected from and the sentence example. 4. The accented prosodic reading character string generation unit (4) is a sentence example selected from the sentence example table unit (2) following the proper noun output from the Kana reading conversion unit (3). , Begins with a suffix that acts to make the accented side of the proper noun into an O type, such as "nearby",
The speech synthesizer according to claim 1, wherein the accent type of the proper noun is O type. 5. The accented prosodic reading character string generation unit (4) selects a sentence example selected from the sentence example table unit (2) following the proper noun output from the Kana reading conversion unit (3). Is characterized by placing an accent nucleus at the third mora from the end of the proper noun when it does not start with a suffix that acts to make the accent side of the proper noun into an O type, such as "nearby". The speech synthesizer according to claim 1. 6. When the third mora from the ending of the proper noun is either “long sound” or “grandphon”, the accent nucleus is shifted and the accent nucleus is placed at the fourth mora from the ending. The speech synthesizer according to claim 5. 7. The accented prosodic reading character string generation section (4), when the mora length of the proper noun output from the kana-reading conversion section (3) is large, the pause length at a predetermined break in the proper noun. 1 is inserted.
The speech synthesizer according to. 8. The reading character string with accent prosody (4)
Is composed by changing the pause length between the number and the sentence example selected from the sentence example table section (2) depending on the mora length of the numeric code string output from the Kana-yomi conversion unit (3). The speech synthesizing apparatus according to claim 2, wherein 9. The accented prosodic reading character string (4)
Means that the sentence examples selected from the sentence example table unit (2) following the numerical code string output from the Kana-yomi conversion unit (3) are number accents such as “degree” and “preceding”. 3. The speech synthesizer according to claim 2, wherein the accent type of the numeral string is set to O type when it starts with a suffix having an action of making side to O type. 10. The sentence example selected from the sentence example table section (2) of the accented prosodic reading character string (4) following the numeric code string output from the Kana-yomi conversion section (3). Is not beginning with a suffix that has the effect of making the accent side of a number an O-type, such as "degree" or "destination", the 3
The speech synthesis apparatus according to claim 1, wherein an accent nucleus is placed on the mora eyes. 11. When the third mora from the end of the code sequence of the numeral is either “long sound” or “consonant”, the accent nucleus is shifted and the accent nucleus is placed at the fourth mora from the end. The speech synthesizer according to claim 10. 12. When the numerical code string is time,
The accent nucleus is placed at the third mora from the ending of “00” at the time indicated by “00 hour ΔΔ minute”, and the accent nucleus position is set when the third mora from the ending is either “long sound” or “consonant”. 11. The speech synthesis apparatus according to claim 10, wherein the speech is shifted from the end of the word to the fourth mora. 13. When the numerical code string is time,
11. The speech synthesis apparatus according to claim 10, wherein an accent nucleus is placed at "ji", which is the third mora from the ending of "00 hour" at the time indicated by "00 hour ΔΔ minute". 14. When the code string of the numeral includes a decimal point, 3 from the ending of “00.” indicated by “00.ΔΔ”.
11. The speech according to claim 10, wherein an accent nucleus is placed in the mora, and the accent nucleus position is shifted to the fourth mora from the end when the third mora from the end is either “long sound” or “consonant”. Synthesizer. 15. When the code string of proper nouns and numbers output from the kana-yomi conversion unit (3) includes a “ga” line sound, it is converted to a nasalized “ga” line sound. The speech synthesizer according to claim 1 or 2. 16. The vowel sequence is lengthened when the code sequence of proper nouns and numbers output from the kana-yomi conversion unit (3) includes the same vowel sequence. Or the speech synthesizer according to 2. 17. When a vowel continuation of “EI” is included in a code sequence of proper nouns and numbers output from the kana-yomi conversion unit (3), this vowel continuation is the same as the same vowel continuation of “EE”. The speech synthesizer according to claim 1 or 2, characterized in that the sound is made longer. 18. When the vowel continuation of “OU” is included in the code sequence of proper nouns and numbers output from the kana-yomi conversion unit (3), this vowel continuation is treated as the same vowel continuation of “OO”. The speech synthesizer according to claim 1 or 2, wherein the speech synthesizer produces a long sound.