JP4311710B2 - Speech synthesis controller - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a controller which facilitates generation/correction of voice synthesis data. <P>SOLUTION: In addition to a voice synthesis part 4, a voice synthesis controller 4 is provided. The voice synthesis controller 4 displays parameters on a screen 6 so that they are easy to intuitively understand. An operator obtains desired voice synthesis data by changing parameters by operating the screen display. <P>COPYRIGHT: (C)2004,JPO&amp;NCIPI

Description

[0001]
BACKGROUND OF THE INVENTION
In a device that performs speech synthesis based on text data, the given text data is subjected to morphological analysis, and then accentuated for each word to obtain speech synthesis data (Patent Document 1). However, the morphological analysis is not necessarily performed accurately, and the accent of each word is not always accurately given.
[0002]
This problem can be solved to some extent by registering many words in the dictionary of the speech synthesis engine or registering words with different accents. However, this method causes another problem that the dictionary is enlarged.
[0003]
Also, the task of obtaining desired speech synthesis data by giving appropriate parameters to the speech synthesis engine requires specialized knowledge of speech and is not easy.
[0004]
Therefore, the present invention seeks to solve the above-described problems by facilitating correction of speech synthesis data.
[0005]
[Patent Document 1]
JP 2003-005776 A
SUMMARY OF THE INVENTION AND EFFECT
(1) A speech synthesis control device according to the present invention is a speech synthesis control device for taking an interface to a speech synthesis unit,
Upon receiving the speech synthesis command, the given character string is given to the speech synthesizer, and the speech synthesis data corresponding to the character string and the parameters used in speech synthesis are acquired from the speech synthesizer, and based on the parameters When the modification of the character string is corrected and a speech synthesis command is given by the operator, the parameter is modified based on the modified modification and given to the speech synthesizer. The voice synthesis data corrected is obtained from the voice synthesis unit.
[0006]
Therefore, the operator can correct desired parameters by using the parameters created by the speech synthesis unit and obtain desired speech synthesis data. In addition, the character string is displayed in a modified form based on the parameter, and the parameter can be modified by modifying the modification, so that the modification operation is easy.
[0007]
(3) The speech synthesis control device according to the present invention is characterized in that the character string is a kanji character string or a kana character string.
[0008]
Therefore, the kanji character string or kana character string is displayed with the modification based on the parameters.
[0009]
(4) In the speech synthesis control device according to the present invention, the character string given from the operator is a kanji character string, and the speech synthesizer generates a kana character string corresponding to the given kanji character string. The character string received from the speech synthesizer is displayed with the modification applied.
[0010]
Therefore, if a kanji character string is input, a corresponding kana character string is generated, and a kana character string with a modification is displayed.
[0011]
(5) An interface program according to the present invention is an interface program for realizing an interface to a speech synthesis unit using a computer,
Upon receiving the voice synthesis command, the given kanji character string is given to the voice synthesizer, and from the voice synthesizer, the voice synthesis data corresponding to the kanji character string, the kana character string corresponding to the kanji character string, and The parameter used in the speech synthesis is acquired, the kana character string is modified and displayed based on the parameter, and the modification of the kana character string is corrected by the operator, and a speech synthesis command is given. Based on the modified modification, the parameter is corrected and given to the speech synthesizer, and the computer is caused to perform processing for acquiring the modified speech synthesis data from the speech synthesizer.
[0012]
Therefore, the operator can correct desired parameters by using the parameters created by the speech synthesis unit and obtain desired speech synthesis data. In addition, the character string is displayed in a modified form based on the parameter, and the parameter can be modified by modifying the modification, so that the modification operation is easy. Furthermore, if a Kanji character string is input, a corresponding kana character string is generated, and a modified kana character string is displayed.
[0013]
(6) The interface program according to the present invention, when the kana character string displayed by the operator is corrected and a voice synthesis command is given, gives the corrected kana character string to the voice synthesizer, and the voice synthesizer It is characterized by acquiring the speech synthesis data corrected from the above.
[0014]
Therefore, when the kana character string generated by the speech synthesizer is incorrect, the operator can correct it and obtain speech synthesis data.
[0015]
(7) The speech synthesis control device according to the present invention is characterized in that the parameter is a parameter related to a length of a sound corresponding to a character.
[0016]
Therefore, the length of the sound can be intuitively recognized by the modification of the characters, and the correction is easy.
[0017]
(8) The speech synthesis control device according to the present invention is characterized in that the parameter is a parameter related to an accent.
[0018]
Therefore, the parameters regarding the accent can be intuitively recognized by the modification of the characters, and the correction is easy.
[0019]
(9) In the speech synthesizer according to the present invention, the parameter is an accent break or morpheme break or both, and the modification of the character string is a modification in which a display break is provided at the position of the break. It is characterized by.
[0020]
Therefore, it is possible to intuitively recognize accent breaks and morpheme breaks by character modification.
[0021]
(10) The speech synthesizer according to the present invention is characterized in that by changing the display partition, the accent partition and morpheme partition as parameters are modified accordingly.
[0022]
Therefore, accent breaks and morpheme breaks can be intuitively recognized by character modification, and correction thereof is easy.
[0023]
(11) In the speech synthesizer according to the present invention, the parameter is the height of the accent, and the modification of the character string is performed by moving the position of each character in a direction perpendicular to the arrangement direction of the character string in accordance with the height of the accent. It is characterized by such modifications.
[0024]
Therefore, the height of the accent can be intuitively recognized by the position of the character.
[0025]
(12) The speech synthesizer according to the present invention is characterized in that, by changing the position of a character, the height of an accent as a parameter is corrected accordingly.
[0026]
Therefore, the height of the accent can be intuitively recognized by the position of the character, and the correction is easy.
[0027]
(13) In the speech synthesizer according to the present invention, the speech synthesizer outputs characteristic information of a plurality of phonemes that are candidates for speech synthesis for each character, and accordingly, a plurality of phoneme candidates for the character When the phoneme segment to be used is changed and a voice synthesis command is given by the operator, the characteristics of the changed phoneme are given to the voice synthesizer, and the voice synthesis data changed from the voice synthesizer is displayed. It is characterized by acquisition.
[0028]
Therefore, the operator can select appropriate phoneme pieces and obtain speech synthesis data.
[0029]
(14) In the speech synthesizer according to the present invention, the characteristics of the phoneme are set to the time length of the phoneme and the phoneme before and after the phoneme in the series of speech data including the phoneme recorded in the speech synthesizer. It is characterized by high and low accents.
[0030]
Therefore, the operator can select an appropriate phoneme piece based on this characteristic.
[0031]
(15) The speech synthesizer according to the present invention is characterized in that speech synthesis data is saved as a speech file in response to a save command.
[0032]
Therefore, the generated speech synthesis data can be saved as a file.
[0033]
(16) The speech synthesizer according to the present invention is characterized in that a character string and a parameter are stored as a speech characteristic file in response to a storage command.
[0034]
Therefore, data necessary for speech synthesis can be stored.
[0035]
(17) The speech synthesizer according to the present invention includes a replacement part that does not determine a specific character string in a part of the character string, and records information for generating a parameter for the replacement part. It is characterized by.
[0036]
Therefore, by inserting a character string for the replacement part, it is possible to obtain a voice characteristic file that can be synthesized with appropriate parameters.
[0037]
(18) The voice characteristic file according to the present invention includes a character string part and a replacement part arranged in the order of speech utterances, and the character string part stores parameters corresponding to the character string and each character. The replacement portion is characterized in that information for determining what parameters are given to each character of the character string when the character string is inserted is recorded.
[0038]
Therefore, speech synthesis can be performed with appropriate parameters by inserting a character string for the replacement part.
[0039]
(19) A program according to the present invention is a program for causing a computer to perform a process of causing a voice synthesis unit to perform voice synthesis based on a voice characteristic file,
The audio file is provided with a character string portion and a replacement portion arranged in the order of speech utterances, and the character string portion includes a character string and parameters corresponding to each character recorded therein. Contains information for determining what parameters are given to each character of the character string when the character string is inserted, and the character string portion is based on the given voice characteristic file. When the operator inserts a character string into the input area and gives a speech synthesis command, the parameter corresponding to the character string part is displayed. Processing for reading out the voice characteristic file and generating parameters for the inserted character string in consideration of the information recorded corresponding to the replacement part, and providing the generated voice synthesis data to the voice synthesis unit Is a program for causing a computer.
[0040]
Therefore, speech synthesis can be performed with appropriate parameters by inserting a character string for the replacement part.
[0041]
(20) The speech synthesizer according to the present invention receives a character string, calculates a parameter, generates speech synthesis data corresponding to the character string based on the character string and the parameter, and converts the character string based on the parameter. When the modification of the character string is corrected by the operator and a speech synthesis command is given, the parameters are modified based on the modified modification to generate speech synthesis data.
[0042]
Therefore, the character string is modified and displayed based on the parameter, and the parameter can be modified by modifying this modification, so that the modification operation is easy.
[0043]
(22) A speech synthesis server device according to the present invention is a speech synthesis server device capable of communicating with a terminal device,
Receives a character string, calculates a parameter, generates speech synthesis data corresponding to the character string based on the character string and the parameter, and modifies the character string based on the parameter to display data for display on the terminal device And when the modification of the character string is corrected by the operator of the terminal device and a speech synthesis command is transmitted, speech synthesis data is generated based on the parameter corresponding to the modified modification, and the terminal device Send to.
[0044]
Therefore, the character string can be modified and displayed on the terminal device based on the parameter, and the modified voice synthesis data can be generated in the server device by modifying this modification, so that the modification operation is easy.
[0045]
(24) The server device according to the present invention transmits a voice characteristic file to the terminal device in response to a request from the terminal device,
The voice characteristic file includes a character string portion and a replacement portion arranged in the order of speech utterances. The character string portion stores a character string and parameters corresponding to each character, and the replacement portion includes Information for determining what parameters are given to each character of the character string when the character string is inserted is recorded.
[0046]
Therefore, the voice characteristic file can be transmitted to the terminal device.
[0047]
(20) A speech synthesis method according to the present invention receives a character string, calculates a parameter, generates speech synthesis data corresponding to the character string based on the character string and the parameter, and converts the character string based on the parameter. When the modification of the character string is corrected by the operator and a speech synthesis command is given, the parameters are modified based on the modified modification to generate speech synthesis data.
[0048]
Therefore, the character string is modified and displayed based on the parameter, and the parameter can be modified by modifying this modification, so that the modification operation is easy.
[0049]
In this invention, “modification of character string” is to change the visual attributes such as size, color, and arrangement of characters constituting the character string so that they can be recognized in comparison with other characters. Say.
[0050]
The “kanji magic character string” refers to a character string including at least one kanji character.
[0051]
The “parameter” is characteristic data necessary for speech synthesis using a character string, and includes, for example, accent height, accent position, speech length, and the like.
[0052]
The “program” is a concept that includes not only a program that can be directly executed by the CPU, but also a source format program, a compressed program, an encrypted program, and the like.
[0053]
DETAILED DESCRIPTION OF THE INVENTION
1. First embodiment
(1) Overall configuration and overview
FIG. 1 shows the overall configuration of a speech synthesis control device and speech synthesis unit according to an embodiment of the present invention. The speech synthesizer 4 creates speech synthesis data by giving a character string. When the speech synthesizer 4 gives a character string and does not give a parameter for speech synthesis, the speech synthesizer 4 generates a parameter by itself and creates speech synthesis data. When a character string is given together with a parameter, speech synthesis data is created according to the parameter.
[0054]
The speech synthesis control device 2 is a device for taking an interface with the speech synthesis unit 4. When a character string is given from the speech synthesis control device 2, the speech synthesis unit 4 returns speech synthesis data and its parameters to the speech synthesis control device 2. The speech synthesis control device 2 displays the character string modified with this parameter (see display screen 6). On the screen 6 of FIG. 1, for example, kana characters are displayed so that their positions rise and fall according to the height of the accent.
[0055]
When the operator hears a sound based on the speech synthesis data and feels that the sound is not a desired sound, the operator can change the parameter and change the parameter to change the displayed character string and redo the speech synthesis. . In this way, the operator can obtain desired speech synthesis data.
[0056]
(2) Hardware configuration
FIG. 2 shows a hardware configuration when implemented using a CPU. Connected to the CPU 10 are a display 12, a memory 14, a mouse / keyboard 16, a hard disk (recording device) 18, a CD-ROM drive 20, and a sound card 22.
[0057]
In addition to an operating system (not shown) such as WINDOWS (trademark), the hard disk 18 stores a speech synthesis engine 28 (speech synthesis program), its dictionary 30, and an interface program 26. These programs and data are those recorded in the CD-ROM 32 and installed in the hard disk 18 via the CD-ROM 20. Note that the speech synthesis engine 28 and the interface program 26 exhibit their functions in cooperation with the operating system. For details of the speech synthesis engine 28, see, for example, Japanese Patent No. 32020163 by the present applicant.
[0058]
The sound card 22 converts the given voice synthesis data into an analog waveform and outputs it to the speaker 24.
[0059]
(3) Interface program processing
A flowchart of the interface program 26 recorded on the hard disk 18 is shown in FIG.
[0060]
First, in step S1, the CPU 10 displays an initial screen shown in FIG. The operator uses the keyboard 16 to input a kanji character string into the text input area 40. FIG. 8 shows a state in which the operator inputs “Arashima Town Road”.
[0061]
Next, when the operator clicks the voice synthesis command button 42 with the mouse 16 and a voice synthesis command is given (step S2), the CPU 10 outputs the input kanji character string to the voice synthesis engine 28. (Step S3).
[0062]
In response to this, the speech synthesis engine 28 returns parameters for speech synthesis to the CPU 10 (step S3). The CPU 10 temporarily stores this parameter in the memory 14. Here, the parameter is information necessary for speech synthesis, and is, for example, accent height for each character, accent segmentation, morpheme segmentation, or the like. A character string to be given is also a parameter.
[0063]
Some of the parameters are shown in FIG. Based on the given Kanji character string “Arajimacho no Michi”, the reading “Ankara Tocho no Michi” (kana character string) is generated. Also, morpheme separation information 46 is shown by morpheme analysis. Also, the part of speech information 48 is shown for each morpheme. Further, accent delimiter information 50 representing a group of accents is also shown. The accent delimiter information 50 also serves as morpheme delimiter information 46. In addition, accent height information 52 (HL) is shown for each kana character.
[0064]
CPU10 which received the parameter modifies a character string based on this parameter, and displays it (step S4). As shown in FIG. 10, this display is displayed in the parameter display field 54. The Kana character string “Antarakuchocho no Michi” is displayed with its position up and down based on the accent height information 52. That is, a high accent (for example, “n”) is displayed above, and a low accent (for example, “a”) is displayed below.
[0065]
Further, the kanji magic character string “Arajimacho no Michi” is displayed in a distinguishable manner by being surrounded by a morpheme frame 56 for each morpheme based on the morpheme division information 46 and the accent division information 50.
[0066]
Further, based on the accent delimiter information 50, both the kana character string and the kanji character string are surrounded by the accent frame 58, and the accent delimiter is clearly shown.
[0067]
Next, the CPU 10 determines which command is given in step S5. Here, since the voice synthesis command is given, the process proceeds to the voice synthesis process in step S6.
[0068]
A flowchart of the speech synthesis process is shown in FIG. The CPU 10 gives the parameters (see FIG. 9) currently stored in the memory 14 to the speech synthesis engine 28 (step S61). In response to this, the speech synthesis engine 28 generates speech synthesis data and returns it to the CPU 10. The CPU 10 temporarily stores this voice synthesis data in the hard disk 18 (step S62).
[0069]
Next, the CPU 10 gives the voice synthesis data to the sound card 22 (step S63). As a result, sound is output from the speaker 24. The CPU 10 returns to step S5 and waits for the next command.
[0070]
The operator listens to the sound from the speaker 24 and, if this is not the desired sound, edits the parameters.
[0071]
FIG. 5 shows a flowchart of the editing process. Here, first, the case of changing the reading will be described. For example, suppose that “Arashimacho no Michi” is the correct reading for “Antarakucho no Michi” indicated by the speech synthesis engine 28.
[0072]
In this case, the operator first selects a Chinese character whose reading is to be corrected on the screen of FIG. Here, “Easy”, “Island” and “Town” are selected by clicking with the mouse 16. Next, editing of reading is selected from the edit menu 60. As a result, the CPU 10 displays a reading editing screen as shown in FIG. 11 (step S92). The operator inputs a correct reading (kana character string) in the reading input field 62. In the figure, the correct reading "Arashicho" is entered. At this time, the correct part of speech is selected in the part of speech selection field 64. Here, “place name” is selected.
[0073]
The speech synthesis engine 28 can appropriately select an accent of the morpheme itself based on the part of speech of the morpheme, or can determine an appropriate accent based on the relationship with the part of speech of the preceding and following morphemes. Therefore, giving a correct part of speech for a morpheme is important for obtaining a desired synthesized sound.
[0074]
When the editing end button 66 is clicked by the operator, the CPU 10 corrects the parameter based on the editing content (step S96). That is, the parameters stored in the memory 14 are corrected as shown in FIG. At this time, since the reading is changed, the height information of the original accent cannot be used. Therefore, the CPU 10 gives a flat plate-type accent that is the most common for the height of the accent. That is, for “Arashimacho”, only the first character “A” has a low accent, and for the second and subsequent characters “Rashimacho”, the accent is high.
[0075]
Based on the edited parameters, the CPU 10 performs a display in which the character string is modified as shown in FIG. As can be seen from the figure, it can be visually confirmed that the reading, the height of the accent, and the morpheme division have been changed.
[0076]
Thereafter, the CPU 10 returns to step S5 in FIG. 3 and waits for the next command. When the operator clicks the playback button 42 on the screen of FIG. 13 and a voice synthesis command is given, the CPU 10 executes the voice synthesis process of step S6. That is, the parameters of FIG. 12 stored in the memory 14 are given to the speech synthesis engine 28. The CPU 10 receives voice synthesis data from the voice synthesis engine 28 and temporarily stores it in the hard disk 18. Further, the CPU 10 gives the voice synthesis data to the sound card 22 and outputs it as sound (see FIG. 4). Thereafter, the CPU 10 returns to step S5 in FIG. 3 and waits for the next command.
[0077]
The operator hears this sound and determines whether or not the sound is a desired sound. If the desired sound is obtained, the save button 70 is clicked. In response to this, the CPU 10 executes the speech synthesis data storage process of step S7 in FIG.
[0078]
A flowchart of the speech synthesis data storage process is shown in FIG. The CPU 10 records the voice synthesis data temporarily stored in the hard disk 18 as a voice synthesis data file (step S71). In this way, a desired sound can be stored. Note that the voice synthesis data file can be recorded on a portable recording medium such as a flexible disk, or can be transmitted as an attachment to an e-mail or the like.
[0079]
On the other hand, if the desired sound is not obtained, edits such as accent segmentation, morpheme segmentation, accent height and phoneme segment are performed (see FIG. 5).
[0080]
The operator can edit accent breaks as follows (step S43). For example, the case where the accent phrase “Arashima Chono” and the accent phrase “Michi” are combined into one accent phrase on the display screen of FIG. 13 will be described. First, the “Arashima Chono” accent frame 58 and the “Michi” accent frame 58 are selected by the mouse 16. In this state, the accent phrase combination is selected from the edit menu 60. In response to this, the CPU 10 combines the selected “Arashima Chono” and “Michi” into one accent phrase “Arashima Cho no Michi”.
[0081]
The CPU 10 corrects the parameters stored in the memory 14 as shown in FIG. The accent division between “no” and “michi” has been changed to a morpheme division. Therefore, the CPU 10 displays the entire “Arashima Cho no Michi” surrounded by one accent frame 58 (step S96). The accent height information 52 and the part-of-speech information 48 are the same as before.
[0082]
In the above description, the combination of accent phrases has been described. However, one accent phrase can be edited so as to be divided into two accent phrases. In this case, the operator specifies an accent phrase to be divided by the mouse 16 and then selects accent phrase division from the edit menu 60. Furthermore, the accent phrase can be divided by designating the position to be divided with the mouse 16. Also in this case, the corrected parameter is stored in the memory 14 and the display is corrected (step S96).
[0083]
Combination and division of morpheme phrases can be performed in the same manner as described above. Also in this case, the corrected parameter is stored in the memory 14 and the display is corrected (step S96).
[0084]
Edit accent height as follows. The operator uses the mouse 16 to select an accent phrase to be edited with accent height. Next, when the operator selects accent height editing from the edit menu 60, the CPU 10 displays an edit screen for accent height as shown in FIG. In FIG. 15, each kana character is arranged at the vertical position in association with the height of the accent. In this figure, the kana characters other than the kana character “A” are all given high accents.
[0085]
Here, if the accent of the kana character “MA” is to be lowered, the mouse 16 is operated and the kana character frame 72 is dragged downward (step S94). In response to this, the CPU 10 changes the parameters of the memory 14 so as to lower the accent of the kana character “ma”. Further, as shown in FIG. 16, the kana character “MA” kana character frame 72 is moved downward and displayed (step S96). In this way, the accent height can be edited.
[0086]
Note that the speech synthesis engine 28 selects and uses an appropriate sound from many sample speeches recorded in the speech dictionary 30 at the time of speech synthesis. That is, one phoneme piece is selected from a plurality of phoneme pieces for one kana character. The speech synthesis engine 28 returns not only the characteristic information of the selected phoneme piece but also the characteristic information of other phoneme pieces that have been selected but not selected as parameters to the CPU 10 (see step S3).
[0087]
Therefore, in this embodiment, this phoneme piece can be changed. The phoneme is edited as follows. On the accent height editing screen of FIG. 16, the operator double-clicks with the mouse 16 the character frame 72 for which the phoneme segment is to be edited. For example, when the character frame 72 of “Sh” is double-clicked, the CPU 10 displays a screen as shown in FIG.
[0088]
In the figure, a phoneme segment candidate field 90 is displayed below the kana character “shi”, and the characteristics of five phonemes are shown. On the left end of the phoneme piece candidate field 90, codes (numbers) 1 to 5 for specifying phoneme pieces are shown. The characteristics of each phoneme piece are shown according to the rules shown in FIG. “2M” and “a” immediately to the right of the code “1” indicate a phoneme segment immediately before the phoneme segment in the sample sound recorded in the dictionary 30. “3M” and “shi” on the right side indicate the phoneme segment. “4M” and “b” at the right end indicate phonemes immediately after the phoneme in the sample sound recorded in the dictionary 30.
[0089]
The numbers “2”, “3”, and “4” at the beginning of “2M”, “3M”, and “4M” indicate mora positions. “M” indicates the height of the accent in the sample sound. “H” is a high accent, “L” is a low accent, and “M” is an intermediate accent.
[0090]
“A”, “shi”, and “b” are notation of phonemes. That is, the immediately preceding sound in the sample sound is “a”, and the immediately following sound is “b”.
[0091]
The operator looks at the characteristics of each phoneme candidate written in this way, and selects a desired phoneme piece with the mouse 16 (step S95). The CPU 10 stores the code of the selected phoneme piece as a parameter of the memory 14.
[0092]
When the speech synthesis button 42 is clicked after editing as described above, the CPU 10 gives the edited parameters to the speech synthesis engine 28. Therefore, it is possible to obtain voice synthesis data generated with the edited parameters and listen to the sound from the speaker 24.
[0093]
If the desired sound is obtained by editing, the voice synthesis data save button 70 can be clicked to record the voice synthesis data file on the hard disk 18.
[0094]
In this embodiment, the voice synthesis data can be saved as a template (voice characteristic file) instead of being saved as it is. In the template, characters are not specified in a part of a character string, and characters are specified and used at the time of speech synthesis. This is effective when only the name part is changed, such as a lost call call, and the other part can be used repeatedly.
[0095]
When the operator clicks the template button 92 on the screen of FIG. 13, the CPU 10 executes the template data storage process in step S8 of FIG.
[0096]
A flowchart of the template data storage process is shown in FIG. First, the CPU 10 displays a template editing screen on the display 12 as shown in FIG. The operator uses the mouse 16 to specify the morpheme of the part to be used by replacing the character string. Here, it is assumed that the frame 56 of “Arajima Island” is designated. In response to this, the CPU 10 performs display for part-of-speech selection as shown in FIG. The operator selects the part of speech of the character string to be put in the “Arajimacho” part (step S81). Here, for example, it is assumed that a place name is selected. The CPU 10 performs display as shown in FIG.
[0097]
When the operator clicks the save button 94 (see FIG. 19), the CPU 10 reads out the parameters stored in the memory 14 (step S82). Based on this parameter, the CPU 10 generates template data as shown in FIG.
[0098]
The “no” and “michi” portions are character string portions in which specific character strings are designated. The part of ($ place name) is a replacement part into which a character string is inserted by replacement at the time of use. In the replacement part, a specific character string is not specified, and its part of speech is specified. By specifying the part of speech, it is possible to accurately determine parameters such as the height of an appropriate accent according to the relationship with the preceding and following parts of speech during speech synthesis.
[0099]
The CPU 10 records the generated data in FIG. 22 on the hard disk 8 as template data. The template data can be recorded on a portable recording medium such as a flexible disk, or can be transmitted as an attachment to an e-mail or the like.
[0100]
In the above embodiment, a kanji character string is given in step S1, but a kana character string may be given.
[0101]
Moreover, although the example which produces | generates the template which includes a replacement part in one part was shown in the said embodiment, you may make it produce | generate and record the audio | voice characteristic data that all are character string parts.
[0102]
In the above embodiment, the voice synthesis control device 2 and the voice synthesis unit 4 are separated from each other. However, a voice synthesis device in which both are integrated may be used.
[0103]
2. Second embodiment
Next, a template processing program for performing speech synthesis based on the template data will be described. The hardware configuration is the same as in FIG. However, the hard disk 18 stores a template processing program instead of the interface program.
[0104]
A flowchart of the template processing program is shown in FIG. First, the CPU 10 reads template data and displays an editing screen on the display 12 (step S101). FIG. 24 shows a display example of the editing screen when the template data of FIG. 22 is read. The replacement portion of the template data is displayed as the character string input unit 120, and the character string is displayed as the character string portion. Note that the part of speech of the replacement part is displayed below the character string input unit 120. This is for input guidance to the operator.
[0105]
The operator inputs a desired character string into the character string input unit 120 using the keyboard 16. In this case, it is assumed that “Minoh” has been input. When the input is completed and a voice synthesis command button (not shown) is clicked, the CPU 10 determines parameters such as accent height and reading for the character string input to the character string input unit 120 (step S103). At this time, parameters such as reading and accent are determined in consideration of part-of-speech information (place name here) given to the replacement part.
[0106]
Next, the CPU 10 gives parameters to the speech synthesis engine 28 to obtain speech synthesis data of “Esaka no Michi” (step S104). Further, the CPU 104 gives this voice synthesis data to the sound card 22 to obtain a voice output (step S105). Note that this speech synthesis data can also be saved.
[0107]
As described above, if a template is used, the character string in the replacement part can be changed while maintaining the quality of the synthesized speech.
[0108]
In this embodiment, part-of-speech information is used as information for determining parameters, but rules for determining parameters may be used.
[0109]
3. Third embodiment
FIG. 25 shows an embodiment in which the voice synthesis control device 2 and the voice synthesis unit 4 are operated by the server device 204. The terminal device 200 can access the server device 204 via the Internet 202. The hardware configurations of the terminal device 200 and the server device 204 are the same as those in FIG. The server device 204 also stores a template processing program.
[0110]
The terminal device 200 stores a browser program, and information from the server device 204 can be displayed by the browser program. The operator can obtain speech synthesis data corresponding to the character string by accessing the server device 204 and giving a character string. It is also possible to edit parameters.
[0111]
FIG. 26 and FIG. 27 show flowcharts of processing in which the server device 204 generates and downloads speech synthesis data in response to a request from the terminal device 200. In this flowchart, the process of the speech synthesis engine 28 and the process of the interface program 26 are shown without being distinguished.
[0112]
In step S <b> 101, the terminal device 200 requests an input screen from the server device 204. In response to this, the server apparatus 204 transmits an input screen for speech synthesis (step S201). The terminal device 200 displays this input screen (step S102).
[0113]
The operator of the terminal device 200 inputs a character string desired for speech synthesis on the input screen. A screen in which a character string is input is shown in FIG. In this screen, when the operator of the terminal device 200 clicks the voice synthesis command button 42, the voice synthesis command is transmitted to the server device 204 (step S104).
[0114]
The server device 204 generates a parameter based on the input character string and performs speech synthesis (step S202). The server device 204 transmits the speech synthesis data to the terminal device 200. Further, the server device 204 transmits a screen to be displayed by modifying the character string to the terminal device 200 based on the generated parameter (step S203).
[0115]
The terminal device 200 reproduces the voice synthesis data as sound (step S105). Also, the screen sent from the server device 204 is displayed. In this screen, as shown in FIG. 10, the character string is modified by a parameter.
[0116]
If the reproduced sound is not a desired sound, the operator of the terminal device 20 performs editing processing such as reading editing, segment editing, accent editing, and phoneme editing. The correction command by the editing process is transmitted to the server device 204 (step S107).
[0117]
The server device 204 transmits a correction screen in which the position of the character string is corrected based on the correction command to the terminal device 200. Further, the parameter is corrected (step S204). The terminal device 200. The corrected screen is displayed (step S108). For example, a screen as shown in FIG. 13 is displayed.
[0118]
When the operator clicks the voice synthesis command button 42 on this screen, the voice synthesis command is transmitted to the server device 204 (step S109). In response to this, the server device 204 performs speech synthesis based on the corrected parameters (step S205). Furthermore, the speech synthesis data is transmitted to the terminal device 200 (step S206).
[0119]
The terminal device 200 reproduces and outputs the voice synthesis data as sound (step S110). The operator repeats the above editing until a desired sound is obtained.
[0120]
If the desired sound is obtained, the operator clicks the save button 70. Thereby, the terminal device 200 transmits a download request for speech synthesis data to the server device 204 (step S111).
[0121]
In response to this, the server device 204 causes the terminal device 200 to record speech synthesis data (step S207). Thereby, the terminal device 200 can store the speech synthesis data as a file.
[0122]
In the above description, the case of downloading speech synthesis data has been described. The process for downloading a template after synthesizing a desired sound is a flowchart as shown in FIG.
[0123]
When the operator of the terminal device 200 clicks the template creation button 92, a template creation screen request is transmitted to the server device 204 (step S121). In response to this, a template creation screen is sent from the server device 204 (step S211), and the terminal device 200 displays this screen (step S122). This screen is, for example, a screen as shown in FIG.
[0124]
The operator of the terminal device 200 inputs replacement part designation, part of speech designation, and the like (step S123). In this input process, the server device 204 creates a change screen based on the input data, but this is omitted in the flowchart. As a result of the data input, for example, a screen as shown in FIG. 21 is displayed.
[0125]
When the operator of the terminal device 200 clicks the template save button 94, a template download request is transmitted to the server device 204 (step S124). The server device 204 creates a template (step S212), and stores the created template data in the terminal device (step S213). Thereby, in the terminal device 200, template data as shown in FIG. 22 can be preserve | saved (step S125).
[0126]
Speech synthesis data and templates obtained in this way can be distributed to other people via the Internet 202 or the like. Others who have received the voice synthesis data can listen to the synthesized voice if they have the sound card 22. Also, another person who has received the template can access the server device 204 from the terminal device 206 and execute the template processing program to obtain desired speech synthesis data.
[0127]
FIG. 29 shows a processing flowchart for obtaining speech synthesis data based on a template. The operator of the terminal device 206 accesses the server device 204 and transmits a template (step S151). In response to this, the server device 204 transmits a template screen (step S251). For example, a screen as shown in FIG. 24 is transmitted. The terminal device 200 displays this.
[0128]
The operator inputs a desired character string in the replacement part 120 of the template screen (step S152). Further, the operator clicks a voice synthesis command button, and transmits the voice synthesis command to the server device 204 (step S153).
[0129]
In response to this, the server device 204 generates a parameter (step S252) and performs speech synthesis (step S253). Further, the generated speech synthesis data is transmitted to the terminal device 200 (step S254). The terminal device 200 reproduces this voice synthesis data (step S154). In this way, speech synthesis can be performed. Further, the speech synthesis data can be stored in the terminal device.
[0130]
In each of the above embodiments, each function is realized by a program, but a part or all of the functions may be realized by a logic circuit.
[Brief description of the drawings]
FIG. 1 is a diagram showing an overall configuration of a speech synthesis control device and a speech synthesis unit according to an embodiment of the present invention.
FIG. 2 is a diagram showing a hardware configuration when the apparatus of FIG. 1 is realized using a CPU.
FIG. 3 is a flowchart of an interface program.
FIG. 4 is a flowchart of a speech synthesis processing part.
FIG. 5 is a flowchart of an editing process part.
FIG. 6 is a flowchart of storing voice synthesis data.
FIG. 7 is a flowchart of template data storage.
FIG. 8 is an example of an input / work screen.
FIG. 9 is a diagram showing generated parameters.
FIG. 10 is an example of a screen displayed by modifying the form of a character string based on parameters.
FIG. 11 is a diagram showing an editing screen for reading.
FIG. 12 is a diagram showing corrected parameters.
FIG. 13 is a diagram showing a corrected input / work screen.
FIG. 14 is a diagram showing corrected parameters.
FIG. 15 is a diagram showing an accent editing screen.
FIG. 16 is a diagram showing an accent editing screen.
FIG. 17 is a diagram showing a phoneme piece editing screen;
FIG. 18 is a display example of characteristics of phoneme pieces;
FIG. 19 is a diagram showing a template creation screen.
FIG. 20 is a diagram showing a screen for part-of-speech selection.
FIG. 21 is a diagram showing a template creation screen.
FIG. 22 is a diagram showing template data.
FIG. 23 is a flowchart of a template processing program.
FIG. 24 is a screen when performing speech synthesis using a template.
FIG. 25 shows a system configuration when speech synthesis is performed from the terminal device 200 using the server device 204.
FIG. 26 is a flowchart of speech synthesis processing. The terminal device side represents browser program processing, and the server device side represents interface program and speech synthesis engine processing.
FIG. 27 is a flowchart of speech synthesis processing.
FIG. 28 is a flowchart for creating a template.
FIG. 29 is a flowchart showing a reproduction process using a template.
[Explanation of symbols]
2. Speech synthesis control device
4. Speech synthesis unit
6 ... Interface screen

Claims (23)

A speech synthesis control device for interfacing with a speech synthesis unit,
Upon receiving a voice synthesis command, the given character string is given to the voice synthesizer, and from the voice synthesizer, the voice synthesis data corresponding to the character string and the parameters used for the voice synthesis are acquired,
Display a modified string based on the parameters,
When the modification of the character string is corrected by the operator and a speech synthesis command is given, the parameters are corrected based on the modified modification and given to the speech synthesizer. A speech synthesis control device to obtain,
The parameters include accent information and characteristic information of a plurality of phonemes that are candidates for speech synthesis for each character.
In the modification of the character string, the position of each character is moved in a direction perpendicular to the arrangement direction of the character string according to the height of the accent, and a series of speech including a corresponding phoneme piece for each character. In the data, addition of a display of a plurality of phoneme candidate candidates by the height of the accent of the phoneme and the phoneme before and after the phoneme,
When the operator changes the vertical position of each character or changes the phoneme candidate to be used and gives a speech synthesis command, the speech synthesizer displays the changed accent and the characteristics of the changed phoneme candidate. A speech synthesis control device that acquires the synthesized speech synthesis data from the speech synthesis unit. An interface program for realizing an interface to a speech synthesis unit using a computer,
Upon receiving a voice synthesis command, the given character string is given to the voice synthesizer, and from the voice synthesizer, the voice synthesis data corresponding to the character string and the parameters used for the voice synthesis are acquired,
Display a modified string based on the parameters,
When the modification of the character string is corrected by the operator and a speech synthesis command is given, the parameters are corrected based on the modified modification and given to the speech synthesizer. A program for causing a computer to perform acquisition processing ,
The parameters include accent information and characteristic information of a plurality of phonemes that are candidates for speech synthesis for each character.
In the modification of the character string, the position of each character is moved in a direction perpendicular to the arrangement direction of the character string according to the height of the accent, and a series of speech including a corresponding phoneme piece for each character. In the data, addition of a display of a plurality of phoneme candidate candidates by the height of the accent of the phoneme and the phoneme before and after the phoneme,
When the operator changes the vertical position of each character or changes the phoneme candidate to be used and gives a speech synthesis command, the speech synthesizer displays the changed accent and the characteristics of the changed phoneme candidate. A program for causing a computer to perform processing for obtaining changed speech synthesis data from the speech synthesis unit. In the apparatus of claim 1 or the program of claim 2,
The character string is a kanji character string or a kana character string. In the apparatus or program in any one of Claims 1-3,
The character string given by the operator is a character string of kanji.
The speech synthesizer generates a kana character string corresponding to the given kanji character string,
The character string received from the speech synthesizer is displayed with the modification applied. An interface program for realizing an interface to a speech synthesis unit using a computer,
Upon receiving the speech synthesis command, the given kanji character string is given to the speech synthesizer, and from the speech synthesizer, the speech synthesis data corresponding to the kanji character string, the kana character string corresponding to the kanji character string, and Get the parameters used for speech synthesis,
The kana character string is modified and displayed based on the parameter,
When the modification of the kana character string is corrected by the operator and a speech synthesis command is given, the parameters are modified based on the modified modification and given to the speech synthesizer. the process of acquiring a program for causing a computer,
The parameters include accent information and characteristic information of a plurality of phonemes that are candidates for speech synthesis for each character.
In the modification of the character string, the position of each character is moved in a direction perpendicular to the arrangement direction of the character string according to the height of the accent, and a series of speech including a corresponding phoneme piece for each character. In the data, addition of a display of a plurality of phoneme candidate candidates by the height of the accent of the phoneme and the phoneme before and after the phoneme,
When the operator changes the vertical position of each character or changes the phoneme candidate to be used and gives a speech synthesis command, the speech synthesizer displays the changed accent and the characteristics of the changed phoneme candidate. A program for causing a computer to perform processing for obtaining changed speech synthesis data from the speech synthesis unit. In the program of Claim 5,
When the displayed kana character string is corrected by the operator and a voice synthesis command is given, the corrected kana character string is given to the voice synthesizer, and the corrected voice synthesis data is acquired from the voice synthesizer. Features In the apparatus or program in any one of Claims 1-6,
The parameter includes a parameter related to a sound length corresponding to the character. In the apparatus or program in any one of Claims 1-6,
The plurality of phoneme candidate candidates are not displayed until each character is clicked by the user. The apparatus or program according to claim 8.
The parameter is an accent break or morpheme break or both,
The character string is modified such that a delimiter on display is provided at the position of the delimiter. The apparatus or program according to claim 9.
By changing the partition on the display, the accent partition as a parameter is corrected accordingly. In the apparatus or program in any one of Claims 1-10 ,
According to a save command, the voice synthesis data is saved as a voice file. In the apparatus or program in any one of Claims 1-11 ,
Character strings and parameters are saved as voice characteristics files in response to a save command. The apparatus or program of claim 12 ,
A replacement part that does not determine a specific character string is provided in a part of the character string, and information for generating a parameter is recorded for the replacement part. Takes a string and calculates the parameters
Generate speech synthesis data corresponding to the character string based on the character string and parameters,
Display a modified string based on the parameters,
When a modification of the character string is corrected by an operator and a speech synthesis command is given, the speech synthesis control device generates speech synthesis data by correcting parameters based on the modified modification .
The parameters include accent information and characteristic information of a plurality of phonemes that are candidates for speech synthesis for each character.
In the modification of the character string, the position of each character is moved in a direction perpendicular to the arrangement direction of the character string according to the height of the accent, and a series of speech including a corresponding phoneme piece for each character. In the data, addition of a display of a plurality of phoneme candidate candidates by the height of the accent of the phoneme and the phoneme before and after the phoneme,
When the operator changes the vertical position of each character or changes the phoneme candidate to be used and gives a speech synthesis command, the voice is based on the changed accent and the characteristics of the changed phoneme candidate. A speech synthesis control device that generates synthesized data. A program for performing speech synthesis processing using a computer,
Takes a string and calculates the parameters
Generate speech synthesis data corresponding to the character string based on the character string and parameters,
Display a modified string based on the parameters,
When the modification of the character string is corrected by the operator and a speech synthesis command is given, the speech synthesis for causing the computer to perform processing for generating the speech synthesis data by correcting the parameters based on the modified modification A program ,
The parameters include accent information and characteristic information of a plurality of phonemes that are candidates for speech synthesis for each character.
In the modification of the character string, the position of each character is moved in a direction perpendicular to the arrangement direction of the character string according to the height of the accent, and a series of speech including a corresponding phoneme piece for each character. In the data, addition of a display of a plurality of phoneme candidate candidates by the height of the accent of the phoneme and the phoneme before and after the phoneme,
When the operator changes the vertical position of each character or changes the phoneme candidate to be used and gives a speech synthesis command, the voice is based on the changed accent and the characteristics of the changed phoneme candidate. A speech synthesis program for causing a computer to perform processing for generating synthesized data. A speech synthesis server device capable of communicating with a terminal device,
Takes a string and calculates the parameters
Generate speech synthesis data corresponding to the character string based on the character string and parameters,
Sending data for display by modifying the character string based on the parameters to the terminal device,
When the modification of the character string is modified by the operator of the terminal device and a speech synthesis command is transmitted, speech synthesis data is generated based on parameters corresponding to the modified modification and transmitted to the terminal device. A speech synthesis server device ,
The parameters include accent information and characteristic information of a plurality of phonemes that are candidates for speech synthesis for each character.
In the modification of the character string, the position of each character is moved in a direction perpendicular to the arrangement direction of the character string according to the height of the accent, and a series of speech including a corresponding phoneme piece for each character. In the data, addition of a display of a plurality of phoneme candidate candidates by the height of the accent of the phoneme and the phoneme before and after the phoneme,
When the operator changes the vertical position of each character or changes the phoneme candidate to be used and gives a speech synthesis command, the voice is based on the changed accent and the characteristics of the changed phoneme candidate. A speech synthesis server device that generates synthesized data. A program for realizing a speech synthesis server device capable of communicating with a terminal device by a computer,
Takes a string and calculates the parameters
Generate speech synthesis data corresponding to the character string based on the character string and parameters,
Sending data for display by modifying the character string based on the parameters to the terminal device,
When the modification of the character string is modified by the operator of the terminal device and a speech synthesis command is transmitted, speech synthesis data is generated based on parameters corresponding to the modified modification and transmitted to the terminal device. A program for causing a computer to perform processing ,
The parameters include accent information and characteristic information of a plurality of phonemes that are candidates for speech synthesis for each character.
In the modification of the character string, the position of each character is moved in a direction perpendicular to the arrangement direction of the character string according to the height of the accent, and a series of speech including a corresponding phoneme piece for each character. In the data, addition of a display of a plurality of phoneme candidate candidates by the height of the accent of the phoneme and the phoneme before and after the phoneme,
When the operator changes the vertical position of each character or changes the phoneme candidate to be used and gives a speech synthesis command, the voice is based on the changed accent and the characteristics of the changed phoneme candidate. A program for causing a computer to perform processing for generating composite data. In the server apparatus or program of Claim 16 or Claim 17 ,
The server device transmits a voice characteristic file to the terminal device in response to a request from the terminal device.
The voice characteristic file is
It has a string part and a replacement part arranged in the order of speech utterances,
In the character string part, the character string and the parameters corresponding to each character are recorded.
Information for determining what parameters are to be given to each character of the character string when the character string is inserted is recorded in the replacement portion. Takes a string and calculates the parameters
Generate speech synthesis data corresponding to the character string based on the character string and parameters,
Display a modified string based on the parameters,
When the modification of the character string is corrected by an operator and a speech synthesis command is given, the speech synthesis method generates speech synthesis data by correcting parameters based on the modified modification .
The parameters include accent information and characteristic information of a plurality of phonemes that are candidates for speech synthesis for each character.
In the modification of the character string, the position of each character is moved in a direction perpendicular to the arrangement direction of the character string according to the height of the accent, and a series of speech including a corresponding phoneme piece for each character. In the data, addition of a display of a plurality of phoneme candidate candidates by the height of the accent of the phoneme and the phoneme before and after the phoneme,
When the operator changes the vertical position of each character or changes the phoneme candidate to be used and gives a speech synthesis command, the speech synthesizer displays the changed accent and the characteristics of the changed phoneme candidate. A speech synthesis method for obtaining modified speech synthesis data from the speech synthesis unit.   A speech synthesis control device for interfacing with a speech synthesis unit,
  Upon receiving a voice synthesis command, the given character string is given to the voice synthesizer, and from the voice synthesizer, the voice synthesis data corresponding to the character string and the parameters used for the voice synthesis are acquired,
  Display a modified string based on the parameters,
  When the modification of the character string is corrected by the operator and a speech synthesis command is given, the parameters are corrected based on the modified modification and given to the speech synthesizer. A speech synthesis control device to obtain,
  In response to the save command, it is configured to save the character string and parameters as a voice characteristics file,
  A speech synthesis control device, wherein a replacement part that does not determine a specific character string is provided in a part of the character string, and information for generating a parameter is recorded for the replacement part.   An interface program for realizing an interface to a speech synthesis unit using a computer,
  Upon receiving a voice synthesis command, the given character string is given to the voice synthesizer, and from the voice synthesizer, the voice synthesis data corresponding to the character string and the parameters used for the voice synthesis are acquired,
  Display a modified string based on the parameters,
  When the modification of the character string is corrected by the operator and a speech synthesis command is given, the parameters are corrected based on the modified modification and given to the speech synthesizer. A program for causing a computer to perform acquisition processing,
  Processing to save character strings and parameters as voice characteristics files according to the save command, and
  A program for providing a replacement part that does not determine a specific character string in a part of the character string, and causing the computer to perform processing for recording information for generating a parameter for the replacement part.   A speech synthesis server device capable of communicating with a terminal device,
  Takes a string and calculates the parameters
  Generate speech synthesis data corresponding to the character string based on the character string and parameters,
  Sending data for display by modifying the character string based on the parameters to the terminal device,
  When the modification of the character string is modified by the operator of the terminal device and a speech synthesis command is transmitted, speech synthesis data is generated based on parameters corresponding to the modified modification and transmitted to the terminal device. A speech synthesis server device,
  The server device transmits a voice characteristic file to the terminal device in response to a request from the terminal device.
  The voice characteristic file is
  It has a string part and a replacement part arranged in the order of speech utterances,
  In the character string part, the character string and the parameters corresponding to each character are recorded.
  A speech synthesis server device characterized in that information for determining what parameter is to be given to each character of a character string when the character string is inserted is recorded in the replacement part.   A program for realizing a speech synthesis server device capable of communicating with a terminal device by a computer,
  Takes a string and calculates the parameters
  Generate speech synthesis data corresponding to the character string based on the character string and parameters,
  Sending data for display by modifying the character string based on the parameters to the terminal device,
  When the modification of the character string is modified by the operator of the terminal device and a speech synthesis command is transmitted, speech synthesis data is generated based on parameters corresponding to the modified modification and transmitted to the terminal device. A program for causing a computer to perform processing,
  The server device transmits a voice characteristic file to the terminal device in response to a request from the terminal device.
  The voice characteristic file is
  It has a string part and a replacement part arranged in the order of speech utterances,
  In the character string part, the character string and the parameters corresponding to each character are recorded.
  A program in which information for determining what parameter is to be given to each character of a character string when the character string is inserted is recorded in the replacement part.

Images