JPH11231885A - Speech synthesizing device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To assist an audience in understanding the contents of reading by adding a reading speed, a rhythm and an emotional tone to the sound of a synthesizing speech to be read out. SOLUTION: Regarding a speech synthesizing device where a text is linguistically analyzed as a word, a rhythm is generated in the reading of the word, and a speech waveform is generated for the text for synthesizing a speech, an important part specifying part 11 for specifying an important part in the linguistically analyzed text and an important part rhythm part 12 for generating a rhythm for keeping the important part speed by the important part specifying part 11, different from parts other than the important part in the text are provided, together with an important part generation part 13 for generating a speech waveform for the important part.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a speech synthesizing apparatus for synthesizing a text sentence into a voice and reading it out. The present invention relates to a device capable of assisting the understanding of the device.

[0002]

2. Description of the Related Art In a conventional speech synthesizer, a text sentence is linguistically interpreted, and the "solid sentence" of the text sentence is interpreted into words of various parts of speech, and is converted into a phonetic symbol sequence comprising phoneme symbols and prosodic symbols. You. Next, prosody processing such as a duration time of each phoneme, a pitch pattern, a pause length, and intonation is performed on the phonetic symbol string. For example, synthesized speech is synthesized by connecting speech waveform segments using phonemes as synthesis units.

[0003] Such a speech synthesizer is used in the field of general equipment, for example, for communication with a service station that performs data transmission.

[0004]

In recent years, the number of service stations for transmitting various types of data has been increasing. On the other hand, as a recipient of this information, for example, in a car, while driving, the information is displayed on the screen, but cannot be viewed at all times.

[0005] In the conventional speech synthesizer, the speech to be read out is provided with accents and intonations as described above so as to make it difficult to hear. There is a problem that the listener cannot understand unless he listens carefully. In addition, when there is an exclamation point “!” Or a question mark “?” In a text sentence, an emotional expression is performed. However, there is no other emotional expression, and the voice read out is generally emotional. There is a problem that expression is poor.

SUMMARY OF THE INVENTION Accordingly, it is an object of the present invention to provide a speech synthesizer that can draw attention to a listener and can read aloud with an emotional expression.

[0007]

According to the present invention, in order to solve the above-mentioned problems, text is linguistically analyzed as words.
In a speech synthesizer for generating a prosody for reading a word and generating a speech waveform for a text to synthesize a speech, an important part designating unit for designating an important part in a linguistically analyzed text; An important part prosody part that forms a prosody that makes the important part specified by the part different from other parts other than the important part in the text;
An important part generation unit that generates a voice waveform for the important part is provided. By this means, subjects, predicates, numbers, and proper nouns are read out in slow or loud voice relative to other text parts as important parts, making it possible to synthesize speech that is easy for listeners to hear become.

[0008] Further, an emotion word dictionary for storing words rich in emotional expression with respect to a normal word dictionary used for language analysis of text, and a speech waveform of a word stored in the normal word dictionary. An emotional waveform dictionary for storing a voice waveform whose pitch has been changed, wherein the important part designating part specifies an important part in a text using the emotional word dictionary, and the important part prosody is Changing the pitch to the important part specified by the important part designating unit to form a prosody of the emotional expression, and the important part speech generating unit using the emotional waveform dictionary unit to generate a speech waveform of the important part. It is characterized by. By this means, a voice with a high pitch is synthesized with a bright emotional expression and a voice with a low pitch is mixed with a dark emotional expression, so that the listener can hear the emotionally rich reading.

[0009]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a diagram showing a receiver having a speech synthesizer according to the present invention and receiving an input text for speech synthesis from a service station. As shown in the figure, the receiver is provided with an antenna 200 for receiving an audio signal, a receiving unit 201 for processing the received signal into audio, and a speaker 203 for converting the audio signal into audio. The receiver further includes a speech synthesizer 100 that receives various kinds of data from the service station in addition to the speech signal, extracts the data by the receiving unit 201, and inputs the data as a text sentence,
The synthesized speech of the speech synthesis device 100 is output from the speaker 203 via the adding unit 202.

FIG. 2 is a diagram for explaining the speech synthesizer 100 of FIG. The speech synthesis apparatus 100 shown in FIG. 1 includes a text analysis processing unit 1 for inputting a text sentence, and the text analysis processing unit 1 includes a word dictionary unit 2. Word dictionary part 2
Stores a number of words that distinguish parts of speech, such as nouns, verbs, particles, and numbers, and readings and accents of the words. Nouns include many proper nouns. Here, the reading of a word is represented, for example, by a mora corresponding to one character of katakana (or hiragana), and one mora is composed of a consonant which is a phoneme and a vowel.

The text analysis processing section 1 uses the word dictionary section 2 to convert an input "solid sentence" text sentence into words and converts them into phonetic symbol strings in which each word is read and accented. . The speech synthesizer 100 further includes a prosody control processing unit 3 for inputting a phonetic symbol sequence from the text analysis processing unit 1. The prosody control processing unit 3 converts the phonetic symbol sequence into a phoneme sequence, and generates a phoneme sequence for each phoneme. Duration, pitch pattern, pose length, accent, intonation,
Form a prosodic sequence such as amplitude.

The speech synthesizer 100 further includes a speech generation processing unit 4, and the speech generation processing unit 4 includes a waveform dictionary unit 5. The waveform dictionary unit 5 prepares a large amount of speech waveforms of about the phoneme length and stores them as speech waveform segments together with information such as the pitch, amplitude, and duration of phonemes. The speech generation processing unit 4 selects speech waveform segments from the phonetic sequence and the prosodic sequence from the waveform dictionary unit 5 and connects them to form a synthesized speech. The waveform dictionary unit 5 is provided with a normal waveform dictionary unit 5A and important part waveform dictionary units 5B and 5C. Important part waveform dictionary 5
B stores a speech waveform segment having a longer duration time than the normal waveform dictionary unit 5A. The important part waveform dictionary unit 5C stores a voice waveform unit whose amplitude is larger than that of the normal waveform dictionary unit 5A.

The text analysis processing unit 1 further includes an important part designating unit 11 for designating an important part in the input text.
Is provided. The important part designation unit 11 designates, for example, "subject / predicate", "number", and "proper noun" in the text as important parts. The prosody control processing unit 3 includes an important part prosody part 12 for adding a specific prosody to the important part specified by the important part specification part 11. The important part prosody part 12 specifies, for example, a prosody in which only the important part is read “slowly” or “largely”.

The voice synthesizing apparatus 100 includes an important part specifying unit 11.
And a display operation unit 14 for operating the important part prosody unit 12 from outside. FIG. 3 is a diagram illustrating the display operation of the display operation unit 14 in FIG. As shown in the figure, the display operation unit 14 displays buttons for selecting and operating "subject / predicate", "number", and "proper noun" as operation buttons for specifying important parts. "Slow" and "Large" buttons are displayed. In addition to these buttons, a button for “designating a service station” is provided.

First, a specific example of the important part designation unit 11 and the prosody control processing unit 12 will be described. In the display of the display operation 14 in FIG. 3, the operation of the important part specification unit 11 when the “subject / predicate” button is pressed as the specification of the important part will be described below.
explain. FIG. 4 is a diagram showing an example for explaining the operation of the important part designation unit 11 of FIG. As shown on the left side of FIG. 4 (a), the text analysis processing unit 1, for example, "Essentially, the point is important. ] And a text mixed with kanji kana. The text processing unit 1 converts the input text into a phonetic symbol string that is a katakana mora.

The text analysis processing section 1 searches the phonetic symbol string for the subject and the predicate, and the important part designating section 11
Then, “Nanda Kanda de Are, shown on the right side of FIG.
Gist Ha important death. , The noun part excluding the subject particle and the stem part of the predicate are designated. Similarly, in the display of the display operation 14 in FIG. 3, it is assumed that the “number” button has been pressed as the designation of an important part. As shown on the left side of FIG. 4B, the text analysis processing unit 1
For example, "100 km / h. ] And a text mixed with kanji kana. The text processing unit 1 converts the input text into a phonetic symbol string,
The phonetic symbols are retrieved numbers of columns, the important part specification section 11, shown on the right side shown in FIG. 4 (b) "flux one hundred key <br/> b meters Per Hour death. ], The number is specified as underlined.

Similarly, in the display of the display operation 14 in FIG. 3, it is assumed that the button of "proper noun" is pressed as the designation of the important part. As shown on the left side of FIG. 4 (c), the text analysis processing unit 1 outputs, for example, "Soon Kobe. ] And a text mixed with kanji kana. The text processing unit 1 converts the input text into a phonetic symbol string, searches for a proper noun in the phonetic symbol string, and the important part designating unit 11 uses the "mamonak" shown on the right side of FIG. Kobe death. ], The proper noun is specified.

Incidentally, any two of the important parts shown in FIGS. 1A, 1B and 1C may be combined or all may be designated at the same time. The prosody control processing unit 12 specifies “slow” or “large” prosody for the specified important part. In this case, the "slow" and "large" prosody may be specified at the same time.

FIG. 3 shows a case where the service station for the receiver 100 sends a designated label of an important part and a designated label of a prosody to be added thereto to the input text.
Is displayed, the button of "designation of service station" is pressed. FIG. 5 is a diagram showing an example for explaining the operation of the important part designating section 11 of FIG. 2 when the "designation of service station" button is pressed. As shown on the left side of FIG.
In the text analysis processing unit 1, for example, "Whatever the point,"<Big or Slow> key point "is"<Big or Slow> important ". Is entered. in this way,
The important part is specified by the label “”, and the prosody of the specified important part is specified by Big (meaning “big”) or Slow (meaning “slow”) in the label <>.
The important part designating section 11 interprets this label, and
(A) shown in the right-hand side, "Nanda Kanda de array, the main point Ha di
Yuuyou death. , The noun part excluding the subject particle and the stem part of the predicate are extracted and designated as important parts. The important part prosody section 12 interprets the label and extracts and specifies the "slow" or "large" prosody for the important part specified in this way.

FIG. 6 is a diagram for explaining the important part sound generator 13 in the sound generator 4 of FIG. The speech generation processing unit 4 receives the phoneme sequence and the prosody sequence from the prosody control processing unit 3 and outputs the waveform of the phoneme constituting the phoneme sequence to the prosody sequence such as the pitch, amplitude, and duration of the phoneme in the normal state. The connection is made by reading from the waveform dictionary unit 5. For example, a mora “YOUT” which is a phoneme decomposition of a consonant and a vowel of the specified important part “mall”
EN ”, the voice generation processing unit 4 generates the voice waveform shown in FIG. 6A using the normal waveform dictionary unit 5A.

In the important part sound generation unit 13, when the “slow” prosody designation is input from the important part prosody unit 12,
The speech generation processing unit 4 forms a speech waveform that is temporally expanded like a mora “YOUTEN” shown in FIG. 6B by using the important part waveform dictionary unit 5B. In this way, the important part is read out with a slower voice than the other parts, so that the listener can be alerted. In particular, it is easier for elderly people to hear.

When the "large" prosody designation is input from the important part prosody unit 12 to the important part speech generation unit 13, the speech generation processing unit 4 uses the important part waveform dictionary unit 5C as shown in FIG. A voice waveform having a large amplitude is formed like the mora “YOUTEN” shown in FIG. In this way, the important part is read out with a loud voice compared to the other parts, so that the listener can be alerted. In particular, it is easier for elderly people to hear.

Further, it is possible to make the user hear speech synthesis with the nuance intended by the service station that provides the information. FIG. 7 is a diagram illustrating a modification of FIG. The drawing differs from FIG. 1 in the word dictionary 2 and the waveform dictionary unit 5. As the word dictionary 2, a normal word dictionary 2A and an emotion word dictionary 2B are provided. Further, as the waveform dictionary 5, a normal waveform dictionary 5A, an emotional waveform dictionary 5D,
5E.

FIG. 8 is a diagram for explaining a normal word dictionary 2A and an emotion word dictionary 2B provided in the speech synthesizer 100 of FIG. As shown in FIG. 7, as the word dictionary 2, a normal word dictionary 2A and an emotion word dictionary 2B are provided.
As shown in FIG. 8A, ordinary words such as impression, Tokyo, comfort, marriage, retirement, and requiem are registered in the ordinary word dictionary 2A. In the emotion word dictionary 2B, words that specify (1) a bright reading mode and (2) a dark reading mode are identified and registered. For example, words such as happy, fun, and marriage are registered as the bright reading mode (1), and words such as sad and retirement are registered as the dark reading mode (2). As described above, the present invention can be realized by adding the word dictionary for emotion 2B that adds extended information to words that are used as keywords for emotion reading without changing the normal word dictionary 2A.

The emotional waveform dictionary section 5D stores (1) the speech waveform of a word whose pitch is higher than that of the word corresponding to the bright reading mode. The emotional waveform dictionary unit 5E stores (2) a speech waveform of a word whose pitch is lower than that of the word corresponding to the dark reading mode. As shown in FIG. 8B, the normal word dictionary 2 is expanded to add, for example, a symbol (1) that is a mode for reading brightly to the word of marriage and a symbol ( 2) may be added. Compared with the former, it is possible to suppress an increase in the size of the word dictionary.

FIG. 9 is a diagram for explaining an example of display operation when emotional reading of the display operation 14 is performed. A selection as to whether or not to express an emotion is made by the display operation shown in FIG.
When the "yes" button is pressed in the display operation 14, the important part designation unit 11 uses the sentence word dictionary 2B to add the words belonging to the mode (1) that are brightly read to the text sent to the text analysis processing unit 1. Or dark reading mode (2)
It is determined whether a word belonging to is included.

FIG. 10 is a diagram for explaining a reading pitch when the mode (1) for reading brightly is included. As shown in this figure, "Suzuki is married. Because it contains the word marriage corresponding to the mode (1) read aloud bright that marriage to the input text ", the important part designation section 11 this string of phonetic symbols" Suzuki Sanga marriage Shimashi <br/> data. ] Is specified as an important part. The important part prosody unit 12 forms a prosody series in which the pitch is increased with respect to other text parts as shown in FIG. In the important part voice generating unit 13, the emotional waveform dictionary unit 5
A voice is generated by raising the pitch of an important part using D. Increasing the pitch in this way results in a brighter reading.

FIG. 11 is a diagram for explaining the pitch at which a sentence is read out when the mode (2) for reading darkly is included. As shown in this figure, "Suzuki has retired. Contains the word “retirement” corresponding to the mode (2) that reads out darkly as “marriage”.
In 1 the pronunciation symbol string "Suzuki Sanga retirement Shimashi <br/> data. ] Is specified as an important part. The important part prosody unit 12 forms a prosody series for lowering the pitch, as shown in FIG. The important part sound generation unit 13 generates a sound by lowering the pitch for the important part using the emotional waveform dictionary unit 5E. When the pitch is lowered in this way, a dark reading is obtained.

Therefore, it is possible to synthesize a voice rich in emotional expression for the listener. Next, the receiver 1
In the case where the service station for 00 sends a designation label of an important part related to emotion and a designation label of a prosody to be added thereto to the input text, the display operation 14 shown in FIG.
Is displayed, the button "Specify service station" is pressed.

FIG. 12 is a diagram showing an example for explaining the operation of the important part designating section 11 of FIG. 2 when the "designation of service station" button is pressed in the display operation of FIG. FIG. 12 (a)
As shown on the left side of FIG.
"Mr. Suzuki got married <High>. Is entered. As described above, the important part is specified by the label “”, and the prosody of the emotion of the specified important part is specified by High (meaning pitch height) in the label <>.
The important part designating section 11 interprets this label and
(A) shown in the right-hand side "Suzuki Sangha marriage Shimashita. 』
Important parts are designated as underlined. The important part prosody section 12 interprets the label and extracts and specifies a prosody with a high pitch for the important part specified in this way.

Also, as shown on the left side of FIG. 12B, the text analysis processing unit 1 outputs, for example, “Mr.
Retired. " Is entered. As described above, the important part is designated by the label “”, and the prosody of the emotion of the designated important part is set to the low level in the label <>.
(Meaning low pitch). Important part designation section 1
In 1 interprets this label, and FIG. 12 (b) shows the right "Suzuki SANGA retirement Shimashita. Important parts are designated as underlined. The important part prosody section 12 interprets the label and extracts and specifies a prosody with a low pitch for the important part specified in this way.

As described above, it is possible to make the user hear the speech synthesis with the nuance intended by the service station that provides the information.

[0033]

As described above, according to the present invention, as important parts, the subject / predicate, numbers and proper nouns are read aloud slowly or loudly with respect to other text parts. It is possible to synthesize speech that is easy for the listener to hear. Furthermore, since a voice with a high pitch is synthesized for a bright emotional expression and a voice with a low pitch for a dark emotional expression, the listener can hear a rich emotional reading.

[Brief description of the drawings]

FIG. 1 is a diagram showing a receiver having a speech synthesizer according to the present invention and receiving an input text for speech synthesis from a service station.

FIG. 2 is a diagram illustrating the speech synthesizer 100 of FIG.

FIG. 3 is a diagram illustrating a display operation of a display operation unit 14 in FIG.

FIG. 4 is a diagram showing an example for explaining the operation of an important part designation unit 11 of FIG. 2;

FIG. 5 is a diagram showing an example for explaining the operation of the important part designating unit 11 of FIG. 2 when the “designation of service station” button is pressed.

FIG. 6 is a diagram illustrating an important partial sound generation unit 13 in the sound generation processing unit 4 of FIG. 2;

FIG. 7 is a diagram illustrating a modification of FIG. 2;

8 is a diagram illustrating a normal word dictionary 2A and an emotion word dictionary 2B provided in the speech synthesis device 100 of FIG. 7;

FIG. 9 is a diagram illustrating a display operation example when emotional reading of the display operation 14 is performed.

FIG. 10 is a diagram illustrating a reading pitch when a bright reading mode (1) is included.

FIG. 11 is a diagram illustrating a pitch at which a sentence is read out when a dark read mode (2) is included.

12 is a diagram illustrating an example of an operation of the important part designating unit 11 of FIG. 2 when a “designation of service station” button is pressed in the display operation of FIG. 9;

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 ... Text analysis processing part 2, 2A, 2B ... Word dictionary part 3 ... Prosody control processing part 4 ... Speech generation processing part 5, 5A, 5B, 5C, 5D, 5E ... Waveform dictionary part 11 ... Important part designation part 12 ... Important part prosody part 13 important part sound generation part 100 voice synthesis part 200 antenna 201 reception part 202 addition part 203 speaker

Claims (12)

[Claims] 1. A speech synthesizer for linguistically analyzing a text as a word, generating a prosody for reading the word, and generating a speech waveform for the text to synthesize speech. An important part prosody part that forms a prosody that makes the important part specified by the important part specification part different from other parts other than the important part in the text; An audio synthesizing apparatus, comprising: an important part generation unit that generates an audio waveform for a part. 2. The important part designating unit designates a subject / predicate in a text as an important part.
The speech synthesizer according to claim 1. 3. The speech synthesizer according to claim 1, wherein the important part designating unit designates a number in a text as an important part. 4. The speech synthesizer according to claim 1, wherein the important part designation unit designates a proper noun in a text as an important part. 5. The speech synthesizer according to claim 1, wherein the important part prosody forms a prosody in which the important part specified by the important part specifying part is read out with a slow voice. . 6. The speech synthesizer according to claim 1, wherein the important part prosody unit forms a prosody in which the important part specified by the important part specification unit is read out with a large voice. 7. The text transmitted from the service bureau,
When an important part, a label for identifying the important part, and a label to which a prosody is to be added to the important part are included, the important part designating unit searches the text for the important part, specifies the important part, and specifies the important part prosody. The unit searches for a prosody for the important part to form a searched prosody,
The speech synthesizer according to claim 1. 8. An emotional word dictionary for storing words rich in emotional expression with respect to an ordinary word dictionary used for language analysis of text, and a voice of a word stored in the ordinary word dictionary. An emotional waveform dictionary unit for storing a voice waveform having a pitch changed with respect to the waveform, wherein the important part designating unit designates an important part in a text using the emotional word dictionary unit, The unit changes the pitch to the important part designated by the important part designating unit to form a prosody of the emotional expression, and the important part speech generating unit generates the speech waveform of the important part using the emotional waveform dictionary unit. The voice synthesizing apparatus according to claim 1, wherein: 9. The speech synthesizer according to claim 8, wherein the sentence word dictionary unit attaches a label for distinguishing a bright sentiment expression from a dark sentiment expression to a word stored therein. . 10. The important part prosody part according to claim 8, wherein the pitch is increased for a prosody of a bright emotion expression, and is increased for a prosody of a dark emotion expression. Speech synthesizer. 11. The word dictionary unit for emotions is combined with the normal dictionary unit and labels words rich in emotional expression among words used in the normal dictionary unit. A speech synthesizer according to claim 1. 12. When the text transmitted from the service station includes an important part rich in emotional expression and a label for identifying the important part, and a label to add prosody to the important part rich in emotional expression, the important part is included. The specifying unit searches the text for an important part rich in emotional expression and designates an important part rich in emotional expression. The important part prosody unit searches the prosody for the important part rich in emotion and forms the searched prosody. The speech synthesizer according to claim 8, wherein: