JPH11282494A - Speech synthesizer and storage medium - Google Patents

Abstract

(57) [Summary] [Problem] When an accent phrase is an object to which the devoicing rule is applied, a process for devoicing or not devoicing a vowel included in the accent phrase according to the length of the predetermined accent phrase is performed. By doing so, it makes the synthesized sound easier to hear. SOLUTION: In step 4, a Japanese character analysis process is performed while referring to a dictionary 12 to convert to a synthesized character string.
If the vocalization rules stored in the dictionary 12 are applied to the synthesized character string in step 8, and it is determined in step 8 that there is a syllable to which the voicing rule is to be applied among syllables in the accent phrase, At 10, the number of beats n in the accent phrase is counted. Subsequently, if it is determined in step 12 that the number of beats n is smaller than 3, in step 16, phonological data which does not devoice the vowels in the accent phrase is set. Then, the accent, pitch, etc. are set in step 18, and speech synthesis processing is performed in step 20.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a speech synthesizer and a storage medium in which a computer program for causing the speech synthesizer to function is stored. is there.

[0002]

2. Description of the Related Art Conventionally, as the above-mentioned voice synthesizing apparatus, for example, one having the configuration shown in FIG. 3 is known. The speech synthesizer 30 includes words, word reading, accents, grammar,
There is provided a dictionary 12 in which information such as a voiceless rule is registered. Here, the devoicing rule will be described with reference to FIG. 4 which explains a part of the content. The devoicing rule defines when vowels are to be devoiced, and is a rule for obtaining crisp synthesized sounds by devoicing vowels. FIG. 4 shows a part of the devoicing rule. The first line in FIG. 4 describes a part of the devoicing rule that [shi] is devoiced when the next line comes after [shi]. For example, when reading out the end of the sentence “was [de · shi · ta]”, [s
Since the vowel [i] of [hi] is sandwiched between the unvoiced consonants [sh] and [t], the vowel [i] of [shi] is devoiced. Therefore, if the unvoiced rule is used, “was” is read out as [de · sh · ta].

[0003] The speech synthesizer 30 is provided with a phoneme data file 20 in which phoneme data as a basic unit of synthesis is stored. And the Japanese analysis unit 14
The input Japanese text data mixed with Kanji Kana
While referring to the dictionary 12, it is converted into a synthesized character string composed of katakana representing phonemes and prosodic symbols representing prosody information, and the converted synthesized character strings are output to the speech parameter setting unit 32. The prosodic symbols include accent phrases, delimiters indicating phrase phrases, and symbols indicating accents and poses.

A voice parameter setting unit 32 sets a voice unit to be synthesized using the input synthesized character string and the devoicing rule stored in the dictionary 12, and includes a speech unit included in the synthesized character string. Parameters such as the pitch of the synthesized speech and the prosody duration are set based on the accent and accent phrase symbol that are present. For example, "was" above
In the example, the voice unit is set to [de] [sh] [ta]. Subsequently, the voice synthesis unit 34 selects the voice unit set by the voice parameter setting unit 34 from the phoneme data file 20, and synthesizes voice based on the selected voice unit and the set parameters. And
The electric signal output from the voice synthesizer 34 is
4 is converted to voice.

[0005]

However, in the above-described conventional speech synthesizer, the vowel is devoiced by applying the devoicing rule, so that the synthesized voice may be difficult to hear. For example, a two-beat word "seven [shi.
[chi]], the voice synthesis of [shi] and the vowel [i] of [chi] are devoiced, and “sh / ch”
Is read, so the sound is hard to hear.

Accordingly, the present invention provides a method in which a predetermined accent phrase in text data is included in the predetermined accent phrase in accordance with the length of the predetermined accent phrase in a case where the devoicing rule is applied. It is an object of the present invention to realize a voice synthesizer and a storage medium in which vowels are unvoiced or non-voiceless so that synthesized voices can be easily heard.

[0007]

According to the present invention, in order to achieve the above object, according to the first aspect of the present invention, there is provided storage means for storing speech synthesis unit data as a unit for synthesizing speech, First determining means for determining whether or not a vowel included in a predetermined accent phrase in text data is to be devoiced in accordance with a voiceless rule; detecting means for detecting a length of the predetermined accent phrase; A second determining unit that determines whether a length of the accent phrase detected by the detecting unit is longer than a predetermined length, and a determination result by the first determining unit is the predetermined If the result is that the vowel included in the accent phrase is to be devoiced, according to the result of the determination by the second determining means,
Processing means for voiceless or non-voiceless processing of vowels included in the predetermined accent phrase, and speech synthesis from the storage means based on a processing result of the processing means and a determination result by the first determination means. And a synthesizing means for reading the unit data and synthesizing the read speech synthesis unit data.

According to a second aspect of the present invention, in the voice synthesizing apparatus according to the first aspect, the processing unit includes the first processing unit.
Is a determination result that the vowel included in the predetermined accent phrase is to be unvoiced, and the determination result by the second determination unit is that the length of the detected accent phrase is If it is determined that the vowel included in the predetermined accent phrase is unvoiced, the result of the determination by the first determination unit is determined to be longer than the predetermined length. The determination result that the vowel included in the accent phrase is to be devoiced, and the determination result by the second determination unit is that the length of the detected accent phrase is longer than a predetermined length set in advance If the determination result is that there is no vowel, a technical means of performing a process of not devoicing a vowel included in the predetermined accent phrase is adopted.

According to a third aspect of the present invention, in the speech synthesizer according to the first or second aspect, the length of the predetermined accent phrase is the number of beats in the predetermined accent phrase, a voice synthesis unit. A technical measure is adopted that is determined by at least one of the number and the phoneme duration.

According to the fourth aspect of the present invention, a storage area in which speech synthesis unit data serving as a unit for synthesizing speech is stored, and a vowel included in a predetermined accent phrase in the text data in accordance with a de-voicing rule. A first determination program for determining whether or not to make the voice unvoiced, a detection program for detecting the length of the predetermined accent phrase, and the length of the accent phrase detected by the detection program is set in advance. Second to determine whether the length is longer than a predetermined length
In the case where the judgment result of the first judgment program and the judgment result of the first judgment program are the judgment result of devoicing the vowel included in the predetermined accent phrase, according to the judgment result by the second judgment means, A processing program that performs a process of unvoicing or not devoicing a vowel included in the predetermined accent phrase, and performing speech synthesis from the storage unit based on a processing result by the processing program and a determination result by the first determination program. A technical means of a storage medium in which a unit data is read and a data synthesizing program for synthesizing the read voice synthesizing unit data is adopted.

[0011]

According to the first to third aspects of the present invention, the storage means stores speech synthesis unit data which is a unit for synthesizing speech, and the first determination means is adapted to a voiceless rule. Therefore, it is determined whether or not the vowels included in the predetermined accent phrase in the text data are to be devoiced, and the second determination unit determines whether or not the accent phrase detected by the detection unit that detects the length of the predetermined accent phrase is detected. It is determined whether the length is longer than a predetermined length set in advance. If the result of the determination by the first determining means is a result of devoicing a vowel included in the predetermined accent phrase, the processing means: A vowel included in the predetermined accent phrase is unvoiced or a process for not voiceless is performed.
The synthesizing unit reads the speech synthesis unit data from the storage unit based on the processing result by the processing unit and the determination result by the first determination unit, and synthesizes the read speech synthesis unit data. That is, when a predetermined accent phrase in the text data is a target to which the devoicing rule is applied, it is determined whether or not the vowel included in the predetermined accent phrase is devoiced according to the length of the accent phrase. Then, the speech synthesis unit data can be synthesized based on the determination result. Therefore, it is possible to solve the problem that, depending on the length of the accent phrase, it becomes difficult to hear when the vowel is devoiced.

[0012] In particular, in the invention according to claim 2, the processing means determines that the determination result by the first determination means is that the vowel included in the predetermined accent phrase is to be devoiced, and If the result of the determination by the second determining means is that the length of the detected accent phrase is longer than a predetermined length, a vowel included in the predetermined accent phrase is devoiced. Do the processing,
The result of the judgment by the first judging means is a judgment result that the vowel included in the predetermined accent phrase is to be unvoiced, and the result of the judgment by the second judging means is the length of the detected accent phrase. Is not longer than a predetermined length, the vowel included in the predetermined accent phrase is not voiced. In other words, if the accent phrase is shorter (not longer) than a predetermined length, and if the vowels in the accent phrase become unvoiced, it is difficult to hear the synthesized sound. If the accent phrase is longer than the predetermined length, the vowel in the accent phrase is devoiced. Therefore, it is possible to solve the problem that it is difficult to hear when the vowel is devoiced due to the short accent phrase.

The length of the predetermined accent phrase is
As in the third aspect of the present invention, it is preferable to be determined by at least one of the number of beats in the predetermined accent phrase, the number of speech synthesis units, and the phoneme duration.
That is, in the speech synthesizer, the number of beats in a given accent phrase, the number of speech synthesis units, and the duration of the phoneme are used as typical parameters indicating the length of the accent phrase.

According to the fourth aspect of the present invention, a storage area in which speech synthesis unit data serving as a unit for synthesizing speech is stored, and a predetermined accent phrase in text data in accordance with a de-voicing rule. A first determination program for determining whether or not the vowel to be voiced is to be unvoiced, a detection program for detecting the length of the predetermined accent phrase, and a length of the accent phrase detected by the detection program is set in advance. The second determination program that determines whether the length is longer than the predetermined length and the determination result obtained by the first determination program are the determination results that the vowels included in the predetermined accent phrase are to be unvoiced. In the case, the vowel included in the predetermined accent phrase is unvoiced or a process for not unvoiced is performed according to the determination result by the second determination unit. A processing program, a data synthesizing program for reading voice synthesis unit data from the storage means based on a processing result of the processing program and a determination result of the first determination program, and synthesizing the read voice synthesis unit data; Is stored in a storage medium, and the storage medium can be used to implement the invention described in claim 1. That is, the speech synthesizer is controlled by, for example, a CPU incorporated in the speech synthesizer or a computer connected to the speech synthesizer, as described in an embodiment of the invention described later. By providing a storage unit as a storage medium in the speech recognition device, or by installing a computer program stored in the storage medium in a computer, a speech synthesis device and a storage medium in which a synthesized sound can be easily heard can be realized. .

[0015]

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram showing an embodiment of a speech synthesizing apparatus according to the present invention. FIG. 1 is an explanatory diagram showing a schematic configuration of a speech synthesizer of the present embodiment by blocks. The same components as those in the related art are denoted by the same reference numerals, and description thereof is omitted. Speech synthesis device 1 of the present embodiment
0, as shown in FIG. 1, the dictionary 12, the Japanese language analyzing unit 14, the beat counting unit 16, and the voice parameter setting unit 18
, A phoneme data file 20, a voice synthesis unit 22, and a speaker 24. The number-of-beats detection unit 16 counts the syllable corresponding to the katakana in the accent phrase as the number of beats based on the delimiter and the end-of-sentence symbol representing the accent phrase included in the composite character string output from the Japanese analysis unit 14. . The voice parameter setting unit 18 determines the number of beats counted by the number of beats counting unit 16, the voicing rule 12 a stored in the dictionary 12 (see FIG. 4), and the synthesized character string output from the Japanese analysis unit 14. , Appropriate phonological data is set. The speech parameter setting unit 18 sets parameters such as the pitch of the synthesized speech and the prosody duration based on the accents and accent phrases included in the synthesized character string. Subsequently, the voice parameter setting unit 18 converts the set phoneme data and parameters into the voice synthesizing unit 2.
Output to 2.

The speech synthesis unit 22 sequentially reads out the phoneme data output from the speech parameter setting unit 18 from the phoneme data file 20, and reads these read phoneme data based on the parameters output from the speech parameter setting unit 18. The synthesized phoneme data is output to the speaker 24 as an audio signal. And the speaker 24
Converts an input audio signal into audio.

Next, the flow of processing when the speech synthesizer 10 performs speech synthesis will be described with reference to the flowchart of FIG. Note that the dictionary 12 and the phoneme data file 20 are stored in a ROM or HD built in the device.
It is stored in a storage medium such as D (Hard Disk Drive), and a computer program for causing the speech synthesizer 10 to execute the processing shown in FIG. 2 is also stored in the storage medium. Also, here, it is assumed that speech synthesis is performed on the word “seven [shi · chi]” which is a two-beat word, and the word “seven” is an object to be applied to the unvoiced rule. Furthermore, if the number of beats n is 3 or more in the case where the voiceless rule is applied, voicelessness is performed.

First, when the word "seven" is input as text data (step 2), a Japanese analysis process of the word "seven" is performed with reference to the dictionary 12, and the word "seven" is converted into a synthesized character string. (Step 4). Then, step 4
Apply the devoicing rule stored in the dictionary 12 to the synthesized character string converted in step (step 6), and determine whether any syllables in the accent phrase include the syllable to which the devoicing rule is applied. Is determined (step 8). Here, since the word “seven” is a target to which the devoicing rule is applied, the process proceeds to step 10 (step 8: Yes), and the number of beats n of the word “seven” is counted. Subsequently, it is determined whether or not the number of beats n is 3 or more. Since the number of beats n of the word “seven” is 2, the process proceeds to step 16 and the vowel [i] is added to the word “seven”. Unvoiced phoneme data [shi] [chi] is set (step 16).

Then, parameters such as accent, pitch, and phoneme duration included in the synthesized character string converted in step 4 are set (step 18). Subsequently, while reading the phoneme data set in step 16 from the phoneme data file 20, the read phoneme data is
The composition is performed while adding the parameters set in step 18. Here, the synthesis result is [shi · ch
i], and the vowel [i] becomes a voiced crisp synthesized sound. If the number of beats n in the accent phrase to be synthesized is 3 or more (step 12: Ye
s), set phonological data in which the vowels are devoiced (step 14). If it is determined in step 6 that the vowel is not to be unvoiced as a result of the application of the devoicing rule (step 8: No), the vowel is not voiced without detecting the number of beats n. (Step 1
6).

As described above, if the speech synthesizer 10 of the present embodiment is used, even if the devoicing rule is applied, if the number of beats n in the accent phrase is 2 or less,
Since the vowel is not devoiced, it is possible to prevent the vowel from being devoiced and becoming difficult to hear. In the above-described embodiment, the case where the number of beats n in the accent phrase is set to 3 as a criterion for determining whether or not a vowel is to be unvoiced when the voiceless rule is applied has been described as a representative. Other numbers may be used. Further, as the parameter for determining the length of the accent phrase, the number of speech synthesis units, the phoneme duration, and the like can be used in addition to the number of beats.

In the above embodiment, the speech synthesizer 1
0 uses a configuration in which a computer program for executing speech synthesis is stored in a storage medium such as a ROM or an HDD in the apparatus.
The voice synthesizing process may be performed by storing the data in a D-ROM or a floppy disk or the like and installing them using a reading device (not shown) provided in the voice synthesizing device 10. In this case, the CD-ROM or the FD functions as the storage medium according to the fourth aspect. Furthermore, a computer program can be read from an external information processing device via a wired or wireless communication unit and operated.

By the way, the phoneme data file 20 functions as storage means or storage area of the present invention, the beat counting section 16 functions as detection means, and the speech synthesis section 22
Functions as a combining means. Further, the voice parameter setting unit 18 functions as a first determining unit, a second determining unit, and a processing unit of the present invention. Further, the speech synthesizer 10
In the case where the CPU provided in the computer executes the processing shown in FIG. 2, the computer program for executing step 8 corresponds to the first determination program according to claim 4 and executes step 10. Corresponds to the detection program, the computer program for executing step 12 corresponds to the second determination program, and the computer program for executing steps 14 to 18 corresponds to the processing program. , The computer program for executing step 20 corresponds to the data synthesis program.

[0023]

As described above, claims 1 to 3 are as described above.
According to the invention described in (1), storage means for storing speech synthesis unit data which is a unit when synthesizing speech, and whether a vowel included in a predetermined accent phrase in text data is devoiced according to a devoicing rule The first determining means for determining whether or not the accent phrase is detected and the length of the accent phrase detected by the detecting means for detecting the length of the predetermined accent phrase are determined whether or not the length is longer than a predetermined length. If the result of the determination by the second determining means and the result of the first determining means are to determine that the vowel included in the predetermined accent phrase is to be unvoiced, Processing means for performing processing for devoicing or not devoicing vowels included in the predetermined accent phrase, and processing results of the processing means and determination results of the first determination means. A voice synthesizing unit for reading out the voice synthesis unit data from the storage means and synthesizing the read voice synthesis unit data. According to the length of the accent phrase, a process is performed to determine whether or not vowels included in the predetermined accent phrase are to be devoiced, and speech synthesis unit data can be synthesized based on the processing result. Therefore, it is possible to solve the problem that, depending on the length of the accent phrase, it becomes difficult to hear when the vowel is devoiced.

In particular, according to the second aspect of the present invention, the processing unit determines that the first determination unit determines that the vowel included in the predetermined accent phrase is to be unvoiced, and , The determination result by the second determination means is:
If it is determined that the length of the detected accent phrase is longer than a predetermined length, a process for devoicing vowels included in the predetermined accent phrase is performed. Is a result of determining that a vowel included in the predetermined accent phrase is to be devoiced, and the result of the determination by the second determination means is that the length of the detected accent phrase is set in advance. If it is determined that the vowel is not longer than the predetermined length, the vowel included in the predetermined accent phrase is subjected to a process of not devoicing. Can solve the problem that it may be.

According to the third aspect of the present invention, the length of the predetermined accent phrase is at least one of the number of beats, the number of speech synthesis units, and the phoneme duration in the predetermined accent phrase. Therefore, typical parameters can be used in the speech synthesizer to determine the length of the accent phrase.

According to the fourth aspect of the present invention,
A storage area in which speech synthesis unit data serving as a unit for synthesizing speech is stored, and a first judgment is made as to whether or not vowels included in a predetermined accent phrase in text data are to be unvoiced in accordance with a voiceless rule. Determination program, and a detection program for detecting the length of the predetermined accent phrase,
A second determination program that determines whether the length of the accent phrase detected by the detection program is longer than a predetermined length, and a determination result by the first determination program are determined by the first determination program. If the result is that the vowels in the accent phrase
The vowels included in the predetermined accent phrase are devoiced according to the determination result by the second determination unit, or
A speech synthesis unit data is read out from the storage unit based on a processing program for performing a process that does not de-voice, and a processing result by the processing program and a determination result by the first determination program. Since the data synthesizing program to be synthesized and the storage medium in which the data synthesizing program is stored are provided, the storage medium is provided as a storage medium in the voice synthesizing apparatus, or the computer program stored in the storage medium is stored in the voice synthesizing apparatus or The speech synthesizer according to claim 1 can be realized by installing the software in a computer connected to the speech synthesizer.

[Brief description of the drawings]

FIG. 1 is an explanatory diagram showing a schematic configuration of a speech synthesis device according to an embodiment of the present invention by blocks.

FIG. 2 is a flowchart illustrating a processing flow of the speech synthesis device according to the embodiment of the present invention.

FIG. 3 is an explanatory diagram showing a schematic configuration of a conventional voice synthesizing apparatus by blocks.

FIG. 4 is an explanatory diagram showing a part of the contents of a voiceless rule.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 10 Speech synthesizer 12 Dictionary 14 Japanese analysis part 16 Beat count part (detection means) 18 Speech parameter setting part (first and second judgment means, processing means) 20 Phoneme data file (storage means) 22 Speech synthesis part (Synthesizing means) 24 speakers

Claims (4)

[Claims] 1. A storage means for storing speech synthesis unit data as a unit for synthesizing speech, and whether or not vowels included in a predetermined accent phrase in text data are to be devoiced according to a devoicing rule. First determining means for determining the length of the accent phrase, detecting means for detecting the length of the predetermined accent phrase, and whether the length of the accent phrase detected by the detecting means is longer than a predetermined length set in advance A second determining means for determining whether or not the vowel included in the predetermined accent phrase is to be devoiced; Processing means for performing voiceless or non-voiceless processing of vowels included in the predetermined accent phrase in accordance with the determination result by With reading the voice synthesis unit data from said storage means based on a determination result by the stage, and combining means for combining the read-out voice synthesis unit data, speech synthesizer, wherein a provided. 2. The processing unit according to claim 1, wherein the determination result by said first determination unit is a determination result that a vowel included in said predetermined accent phrase is to be unvoiced, and said determination unit determines by said second determination unit. If the result is a determination result that the length of the detected accent phrase is longer than a predetermined length, a process of devoicing vowels included in the predetermined accent phrase is performed. The result of the determination by the first determining means is a result of determining that the vowel included in the predetermined accent phrase is to be unvoiced, and the result of the determination by the second determining means is the length of the detected accent phrase. If the result of the determination is that the vowel included in the predetermined accent phrase is not longer than a predetermined length, a process for not devoicing a vowel included in the predetermined accent phrase is performed. Speech synthesis apparatus according to 1. 3. The length of the predetermined accent phrase is determined by at least one of the number of beats, the number of speech synthesis units, and the phoneme duration in the predetermined accent phrase. The speech synthesizer according to claim 1 or 2. 4. A storage area in which speech synthesis unit data as a unit for synthesizing speech is stored, and whether a vowel included in a predetermined accent phrase in text data is to be devoiced in accordance with a devoicing rule. A determination program for determining the length of the accent phrase, and a detection program for detecting the length of the predetermined accent phrase, whether the length of the accent phrase detected by the detection program is longer than a predetermined length set in advance. A second determination program for determining whether or not a vowel included in the predetermined accent phrase is to be devoiced, if the determination result obtained by the first determination program is a second voice, A processing program for performing a process of devoicing or not devocalizing a vowel included in the predetermined accent phrase according to the determination result by And a data synthesizing program for reading the speech synthesis unit data from the storage means based on the processing result by the program and the judgment result by the first judgment program, and synthesizing the read speech synthesis unit data. A storage medium characterized by the above-mentioned.