JPH06231124A - Voice conversion system - Google Patents

Abstract

PURPOSE:To prepare a document based upon voice by executing operation for preparing the document by voice from the initial stage of the document preparation. CONSTITUTION:In each input of a 'HIRAKANA' (cursive form of Japanese syllabary) character from a character input key 21 or an operation function indicating instruction from one of indication instruction keys 22 to 31, voice data corresponding to the 'HIRAKANA' character or the instruction are generated and voice data corresponding to converte character data are also generated from a speaker 8, so that document preparation can be attained based on the generated voice data. Since document preparation can be attained without observing a display part 5, a visually handicapped person also can prepare the document by means of a word processor.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice conversion system for converting a character into a voice and producing a voice document when the document is produced.

[0002]

2. Description of the Related Art In recent years, with the development of OA equipment, the population of so-called word processors, etc., which convert hiragana into kanji and create documents, has been increasing rapidly, and word processors for various purposes have been developed. ing. For example, a business word processor,
These include personal word processors for individuals and word processor software installed on a personal computer.

As such a word processor for each application, a so-called voice word processor using a voice technique has begun to be developed, and a system for reading out a document created by the word processor using a voice synthesis technique or the like has been developed.

[0004]

However, since the system described above is a system that simply synthesizes the text created by a word processor and reads it aloud, sentences are created while looking at the display. The number of people complaining of tiredness is increasing, and there are problems such as writing sentences without looking at the display, or making it difficult for a visually impaired person to write sentences with a word processor.

The present invention has been made in view of the above circumstances, and provides a voice conversion system capable of uttering an operation for making a document by voice and making a document by voice from an early stage of document production. With the goal.

[0006]

In order to achieve the above object, the present invention provides an input means for inputting a character string composed of a plurality of hiragana characters and an instruction command of an operation function, and a hiragana or instruction command from the input means. A voicing means for uttering voice data corresponding to the hiragana or an instruction command each time it is input, and a character string input from the input means is converted into a plurality of character data candidates including kanji and the converted characters. Character data conversion means for uttering voice data corresponding to the data from the voicing means, character data determination means for deciding one character data from a plurality of character data candidates converted by the character data conversion means, and Converts the character data determined at the time of determination by the character data determination means into the corresponding first voice data candidate and into a plurality of voice data candidates Audio data conversion means for uttering the converted audio data candidates from the voicing means, and audio data determination means for deciding one audio data from a plurality of audio data candidates converted by the audio data conversion means. The present invention is characterized by comprising: storage means for storing the voice data determined by the voice data determination means; and reproduction means for continuously reproducing the voice data stored in the storage means.

Further, the present invention is characterized by having voices of a plurality of speakers as a voice dictionary for the character data.

Further, the present invention is characterized by comprising switching means for switching and uttering voice data by voice synthesis sound or voice registered in advance when voice data is uttered by the voicing means.

Further, the present invention is characterized by comprising fine adjustment means for finely adjusting the voice data and uttering it when converting the character data into the voice data.

Further, the present invention is characterized by comprising a registration means for registering the voice data finely adjusted by the fine adjustment means as voice data in a voice dictionary.

[0011]

Since the present invention is configured as described above, every time the voicing means utters the hiragana or the command corresponding to the hiragana or the command, the voice data corresponding to the hiragana or the command is converted into character data. By uttering corresponding voice data, a document can be created based on the uttered voice data, and even a visually handicapped person can create a document by a word processor.

Further, according to the present invention, since the voice dictionary has the voices of a plurality of speakers for the character data, it is possible to create a voice document with a voice which the operator can easily hear or a voice of his / her preference.

Further, according to the present invention, since the switching means is provided, the flexibility of the system can be expanded by uttering voice data as a voice synthesis voice or a voice registered in advance.

Further, since the present invention is provided with the fine adjustment means, the sound quality of the voice data is improved, thereby preventing misunderstanding, and the voice data having the improved sound quality is registered in the voice dictionary. This makes it easy to create a voice document.

[0015]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a diagram showing a schematic configuration of a voice conversion system according to an embodiment of the present invention. A portion surrounded by a chain double-dashed line shows an ordinary word processing system 1, and the voice conversion system of the present invention. Is constructed by adding elements other than the portion surrounded by the two-dot chain line to the word processor system 1 having the above configuration.

In the figure, the keys input from the character input keys 21 of the keyboard unit 2 are a CPU for controlling the operation of the voice conversion system, a ROM for storing a control program,
And the character information corresponding to the determined key is displayed on the display unit 5 while being discriminated by the control unit 4 including a RAM for temporarily storing the processed data, and voices of a plurality of speakers converted by, for example, the PCM method. The voice data corresponding to the character input key 21 input from the voice dictionary 63 of the storage unit 6 in which voice data such as or voice of the user is registered in advance is read by the control unit 4 under the control of the control unit 4. Is output. The voice conversion LSI 7 converts the voice data of the input digital signal into an analog signal and outputs it to the speaker 8
Is output as a voice from. Also, when an instruction command key such as a character conversion key 22 or a character determination key 23 for inputting an instruction command of an operation function of this system other than the character input key 21 configuring the keyboard unit 2 is input, the character input key
Similar to the input of 21, voice data corresponding to the input instruction command key is output as voice from the speaker 8.

In the case of kana-kanji conversion, the character input key 21
After inputting a character string consisting of multiple hiragana characters from, the control unit 4 causes the control unit 4 to input the character string input from the character dictionary 61 for Kana-Kanji conversion composed of the ROM that constitutes the storage unit 6. As the character data corresponding to, the first kanji candidate is read out from the kanji candidates of characters / words consisting of kanji, katakana, or romaji and displayed on the display unit 5. Simultaneously with the display of the first Kanji candidate, the voice data corresponding to the first Kanji candidate is read from the voice dictionary 63 and output as a voice from the speaker 8. The operator presses the character conversion key 22 based on the voice uttered from the display unit 5 or the speaker 8 until the correct Kanji candidate is displayed or uttered, and presses the character determination key 23 when the correct Kanji candidate is displayed or uttered. Thus, the control unit 4 determines the Kanji candidate as the determined character data, and stores the determined character data in the character document storage unit 62 including the RAM of the storage unit 6.

As described above, by repeating the input of the character string, the kana-kanji conversion, and the determination of the character data,
A desired character document is created, the character document being created is stored in the character document storage unit 52, and the created character document is made up of a floppy disk device (FDD) or a hard disk device (HDD) through a predetermined procedure. Stored in storage device 9.

When creating a voice document, the character determination key 23 is pressed to display the control section 4 on the display section 5,
Alternatively, when the Kanji candidate uttered from the speaker 8 is determined as the determined character data, the first voice candidate of the voice data corresponding to the Kanji candidate determined from the voice dictionary 63 is controlled and read by the control unit 4, It is output to the voice conversion LSI 7. The voice conversion LSI 7 converts the input digital signal voice data into an analog signal and outputs it to the speaker.
Speech output from 8 The operator presses the voice conversion key 24 until the voice with a pronunciation that is easy to hear from the speaker 8, that is, the voice data of the desired pronunciation is produced, and when the desired pronunciation is output, presses the voice determination key 25 to control the control unit 4. The voice data is determined as the determined voice data, and the determined voice data is stored in the voice document storage unit 64 including a RAM.

As described above, by repeating the input of the character string, the kana-kanji conversion, the determination of the character data, the conversion to the voice data, and the determination of the voice data, a desired voice document corresponding to the determined character data is obtained. The voice document being created in association with the text document is being stored in the voice document storage unit 64. The created voice document is stored in the tape recorder of the storage device 9 through a predetermined procedure.

As a method of uttering a voice corresponding to a character, for example, the ASCII code of the input character may be used as the address of the voice dictionary 63 as it is, or the address of the character dictionary 61 and the voice dictionary 63. The addresses may be linked, but since there are multiple voice data for one character, the address obtained by multiplying the address of the character dictionary 61 by several times is used as the address of the voice dictionary 63, so that only the multiple The address area is assigned to one character. It is appropriate that the multiple is the number of voice data assigned to one character.

However, by simply multiplying the address of the character dictionary 61 by several times, the address of the character dictionary 61 and the voice dictionary are
Since the addresses of 63 are overlapped, it is sufficient to divide them by hardware or to bias the addresses of the voice dictionary 63 by the addresses of the character dictionary 61. For example, a multiple of 4
, The address of the character voice dictionary 63 whose address is 5 is 20 + bias value, and the address of the character voice dictionary 63 whose character dictionary 61 is 6 is 24 addresses.
This is a + bias value, and four address areas are given to one character. Therefore, it has an independent quaternary counter as an offset address, and the voice conversion key 24
All addresses in the voice dictionary 63 can be expressed by counting up the offset address each time is pressed, and different voice data may be registered in each of the four addresses. With this configuration, when the voice conversion key 24 is pressed four times or more, the voice data once uttered can be uttered again, and the endless voice with the same pronunciation can be uttered many times. Note that the above description has dealt with the processing of voice data corresponding to one character, but the same can be applied to character data converted into Chinese characters or words.

Similarly to the character dictionary 61, the voice dictionary 63 has information about the voice data determined last time and use frequency information about each voice data, and the voice data output based on these information. Control unit 3 has priority
The audio data of the addresses determined according to the priority order are sequentially output. For example, the audio data determined last time has the highest priority, and then the priority increases in the order of frequency of use. Also, the voice data uttered corresponding to the input of the character input key 21 or the instruction command key is
For example, audio data with the highest priority is output.

The determined character data and voice data are stored in the character document storage unit 62 and the voice document storage unit 64, respectively, and the respective data are stored correspondingly. That is, character data and voice data are stored in different areas of the same address, and the respective areas are stored in the character document storage unit 62.
The voice document storage unit 64 is used to associate character data with voice data. As a method of storing the character data and the voice data in association with each other, a method in which an address obtained by adding a certain number to the storage address of the character data is associated as the storage address of the voice data may be used.

Reference numeral 26 is a reproduction key for reproducing a voice document in order to check the pronunciation of the voice document during or after the production. By pressing the reproduction key 26, the voice is controlled by the control unit 4. The voice document stored in the document storage unit 64 is read and output from the speaker 8 via the voice conversion LSI 7. At this time, under the control of the control unit 4, the character data corresponding to the voice data output from the speaker 8 is displayed on the display unit 5 in a special display such as reverse display, so that the operator can confirm that the converted characters are correct. It is confirmed whether or not there is any.

Reference numeral 27 denotes a fine adjustment key that is pressed when finely adjusting the voice data to a desired sound quality.
When 27 is pressed, the control unit 4 recognizes the pressing of the fine adjustment key 27, and the same audio data as the audio data output to the audio conversion LSI 7 is output by, for example, a DSP (Digital Signal P
output to the sound quality improvement processor 10 composed of a rocessor). Each time the fine adjustment key 27 is pressed, the sound quality improvement processor 10
Fine-tunes and converts the sound quality of the input audio data, and the converted audio data is output to the audio conversion LSI 7 and uttered from the speaker 8. The operator listens to the converted voice from the speaker 8 and presses the voice determination key 25 when the desired sound quality is reached to determine the voice. The voice of the determined sound quality is stored in the voice document storage unit 64.

The sound quality of the audio data converted by the sound quality improvement processor 10 as described above depends on the type of processor used and the conversion method set. In addition, the voice dictionary 63
In order to prepare some voice data with different pronunciations for one character and convert the sound quality by using the sound quality improvement processor 10 only for fine adjustment of the sound quality, or to reduce the capacity of the sound dictionary 63. It is possible to compensate for the lack of voice data in the voice dictionary 63 by reducing the number of different pronunciations for one type of character and significantly changing the voice quality by the tone quality improvement processor 10.

Further, 28 is a mode switching key for switching the operation mode of the voice conversion system, and by pressing this mode switching key 28, the display unit 5 is controlled by the control unit 4.
A mode menu such as registered voice mode, voice synthesis voice mode, voice document creation mode, and voice registration mode is displayed on the display screen of, and the cursor on the display screen is moved by pressing the mode switching key 28 to select the mode menu. As a result, the selected operation mode is switched. Also,
While the mode menu is displayed on the display unit 5, when the cursor moves to the mode menu, the voice data corresponding to the menu information at which the cursor is positioned is uttered, and the voice also determines which mode is selected. .

When the registered sound mode is selected, as described above, the voice data output from the voice dictionary 63 is converted into an analog signal by the voice conversion LSI 7 and the speaker 8 is uttered based on the registered voice data. Is output from
When the voice synthesis voice mode is selected, the voice data output from the voice dictionary 63 is output to the voice synthesis LSI 11 without being output to the voice conversion LSI 7, and the voice synthesis voice code is converted from the voice synthesis LSI 11 to the voice. It is output to the LSI7 and the synthesized voice is output from the speaker 8. Also,
The voice-synthesized sound synthesized by the voice-synthesis LSI 11 is output to the sound quality improvement processor 10 and can be finely adjusted.

When the voice document creation mode is selected, as described above, the voice document is created subsequent to the kana-kanji conversion.

When the voice registration mode is selected, the voice dictionary
It becomes the mode to register the voice data in 63. First, input character data such as characters or kanji corresponding to the voice data to be registered, and press the character determination key 23, and the control unit 4 causes the voice corresponding to the character data in the same manner as the voice data utterance described above. The vacant offset address of the address of the dictionary 63 is read. The operator inputs a voice from the microphone 12 while pressing the voice input key 29. The voice input from the microphone 12 while the voice input key 29 is pressed is
Under the control of 4, the voice conversion LSI 7 converts the voice code to a voice code consisting of a PCM code and registers it in the region of the voice dictionary 63 of the read address. Furthermore, speech synthesis LS
The audio data output from I11 or the audio data whose sound quality has been improved by the sound quality improvement processor 10 is also the voice input key 29.
It is registered in the voice dictionary 53 by making a voice while pressing.

Further, numeral 30 is a cancel key and numeral 31 is a delete key, which is used for canceling or deleting an erroneous character when, for example, an incorrect kana character is input from the character input key 21.

Next, the operation of the speech conversion system of the present invention having the above configuration will be described based on the case of inputting the kanji "kawa".

First, the hiragana character "ka" is input by the character input key 21, the input character "ka" is displayed on the display unit 4, and the voice "ka" is uttered from the speaker 8. The operator confirms that the sound "ka" is generated and presses the character determination key 23. Then, in the same manner, "wa" is input. If input is made by mistake, the cancel key 30 or the delete key 31 is pressed to erase the character input by mistake.
For example, if you mistakenly input “ta”, pressing the cancel key 30 will output the voice “Cancel” from the speaker 8 and confirm that the mistakenly entered “ta” has been canceled. . When deleting with the delete key 31, for example, if “Deleted character” has been deleted is registered in the voice dictionary 63, pressing the delete key 31 causes “” Removed ”and speaker 8
Voiced by. In any case, it is only necessary to convey the role of the key pressed for deletion to the operator, and the spoken word may be any word.

When the character conversion key 22 is pressed at the stage where "KAWA" is displayed in hiragana on the display unit 4, the Kanji corresponding to the character string "KAWA" is read from the character dictionary 51 and converted into Kanji. To be done. Here, for example, it is assumed that the character has been converted to the Chinese character “skin”. In this case, in order to understand what kind of kanji is displayed even if the voice "kawa" is spoken, for example, "skin of mandarin orange" is spoken. In this case as well, it suffices to inform the operator what kind of kanji is displayed, and in short, it is sufficient to register in the voice dictionary 63 words that are easy for the user to understand.

By the way, since the operator does not want to input "skin", the character conversion key 22 is pressed to read the next kanji character candidate. Press and hold the character conversion key 22 until "River" is read out, and "River" is read out. When "Kawasaki-shi" River "" is uttered, press the character decision key 23 to enter "River" into character data. To decide. The “kawa” determined as the character data is stored in the character document storage unit 52. At this time, even if the same "river" cannot be understood by the "kawa" in Kawasaki City, some people can understand it by the "kawa" in Kawagoe City.
It is more appropriate to change the shape of the same word so that it is spoken several times.

Although the operator may finish the document creating operation at this stage and create only a text document, it is possible to create a more accurate document by creating a voice document. The voice document creation mode is selected with the mode switching key 28. When "kawa" is selected as character data by the character determination key 23 by selecting the voice document creation mode, the control unit 4 controls the
The first candidate for the voice data corresponding to "river" is the voice dictionary 63.
Read out from the speaker via the voice conversion LSI7
Spoken from 8. The operator hears the pronunciation produced from the speaker 8, and if the pronunciation is not the desired one, press the voice conversion key 25 to read the voice data of the next candidate,
Output from speaker 8. When the voice with the desired pronunciation is uttered, the voice determination key 25 is pressed to determine the voice data. The determined voice data is stored in the character document storage unit 62.
Is stored in the voice document storage unit 64 which stores voice data corresponding to.

By repeating the input of the character string, the kana-kanji conversion, the determination of the character data, and the determination of the voice data, the desired voice document corresponding to the determined character data is created in association with the character document. , Character document storage unit 62
Is stored in the voice document storage unit 64 in correspondence with.

As described above, by creating a voice document together with a text document, for example, the same "chopstick" has different pronunciations for "bridge" and "chopsticks", and "chopsticks" is pronounced as "bridge". Since there are no people, if the correct pronunciation is registered in the voice dictionary 63, these Chinese characters will not be mixed and the correct characters can be input.

When the reproduction key 26 is pressed after the voice document is created, the voice document stored in the voice document storage unit 64 is read out and the voice is output from the speaker 8 via the voice conversion LSI 7. At this time, character data corresponding to the voice data output from the speaker 8 is displayed on the display unit 5 as a special display with reverse video, and the operator can confirm the pronunciation of the voice document and the conversion of the text document. It is done at the same time.

The "river" uttered from the speaker 8
If you want to change the sound quality by listening to, press the fine adjustment key 27 to adjust the sound quality. That is, the fine adjustment key
When 27 is pressed, the sound data of "River" is output to the sound quality improvement processor 10, and each time the fine adjustment key 27 is pressed, the sound quality improvement processor 10 converts the sound quality of "River". Then, the converted "river" uttered from the speaker 7 is heard, and when the desired sound quality is reached, the sound determination key 25 is pressed to determine the sound quality of the "river". The sound of the determined sound quality is stored in the sound document storage unit 64 as the sound data of “river” converted by the sound quality improvement processor 10.

In addition, at the stage of determining the voice data of "kawa", when the voice data of the desired pronunciation does not exist, the mode switching key 28 is pressed to select the voice registration mode from the operation mode menu. When "kawa" of the character data corresponding to the data is input and the character determination key 23 is pressed, the vacant offset address of the address of the voice dictionary 63 corresponding to "kawa" is read. Next, when the operator inputs the desired pronunciation "kawa" from the microphone 12 while pressing the voice input key 29, the voice conversion LSI
At 7, the voice code is converted into a voice code of "kawa" having a desired pronunciation and registered in the region of the voice dictionary 63 of the read address.

Further, when the voice synthesis voice mode is selected from the operation mode menu, the voice data of "kawa" read from the voice dictionary 63 is output to the voice synthesis LSI 11.
The voice synthesis voice code corresponding to "River" is the voice synthesis LSI11.
Is output to the voice conversion LSI 7, and the synthesized sound of “river” is output from the speaker 8.

In the above embodiment, the voice data registered in the voice dictionary 53 is the voices of a plurality of speakers or one's own voice, but the present invention is not limited to this, and it is difficult to express the natural crying of animals. It is also possible to register various voices in the natural world as voice data in the voice dictionary 53 and select desired voice data. By doing so, a very natural voice document can be created. .

In the above embodiment, the fine adjustment key 27 for exclusive use is used.
I tried to fine-tune the sound quality of the voice data using
The present invention is not limited to this, and fine adjustment may be performed by simultaneously pressing the other two keys, for example, the shift key and the voice conversion key 24. The point is to use other means without providing a dedicated key. Anything can be finely adjusted.

Further, the present invention is not limited to the above-mentioned embodiments, and it goes without saying that various modifications can be made without departing from the gist of the present invention.

[0048]

As described in detail above, according to the voice conversion system of the present invention, every time the voicing means inputs a hiragana or an instruction command from the input means, it produces voice data corresponding to the hiragana or the instruction command. In addition, by uttering voice data corresponding to the converted character data, it becomes possible to create a document based on the uttered voice data. Even free people can create documents by word processing.

Further, it is possible to switch the voice synthesis voice and the voice registered in advance by the switching means, so that the flexibility of the system can be expanded.

Further, since the sound quality of the voice data to be uttered is improved by the fine adjustment means, it becomes easy to listen to the voice document, and it is possible to prevent a mistake in listening.

[Brief description of drawings]

FIG. 1 is a diagram showing a schematic configuration of a voice conversion system according to an embodiment of the present invention.

[Explanation of symbols]

 21 ... Character input key (input means) 22 ... Character conversion key (character data conversion means) 23 ... Character determination key (character data determination means) 24 ... Voice conversion key (voice data conversion means) 25 ... Voice determination key (voice data) 26. Playback key (playback means) 27 ... Fine adjustment key (fine adjustment means) 28 ... Mode switching key (switching means) 29 ... Mode switching key (switching means) 4 ... Control section 62 ... Character document storage section ( Storage means 63 ... Voice dictionary 64 ... Voice document storage unit (storage means) 7 ... Voice conversion LSI (speaking means) 8 ... Speaker (speaking means) 10 ... Sound quality improvement processor (fine adjustment means)

Claims (5)

[Claims] 1. An input means for inputting a character string composed of a plurality of hiragana characters and an instruction command for an operation function, and voice data corresponding to the hiragana or instruction command each time the hiragana or instruction command is input from the input means. Speaking means for uttering and converting the character string input from the inputting means into a plurality of character data candidates including Chinese characters and converting the voice data corresponding to the converted character data from the uttering means Means, a character data determination means for determining one character data from a plurality of character data candidates converted by this character data conversion means, and a character data determined at the time of determination by this character data determination means. Of the voice data and the plurality of voice data candidates, and the converted voice data candidates are uttered from the voicing means. The voice data conversion means for uttering voice, the voice data determination means for determining one voice data from the plurality of voice data candidates converted by the voice data conversion means, and the voice data determined by the voice data determination means An audio conversion system comprising a storage means for storing and a reproduction means for continuously reproducing the audio data stored in the storage means. 2. The voice conversion system according to claim 1, further comprising voices of a plurality of speakers as a voice dictionary for the character data. 3. The voice conversion according to claim 1, further comprising switching means for switching the voice data by voice synthesis voice or a voice registered in advance when the voice data is produced by the voice producing means. system. 4. The voice conversion system according to claim 1, further comprising a fine adjustment means for finely adjusting the voice data and uttering the voice data when converting the character data into the voice data. 5. The voice conversion system according to claim 4, further comprising registration means for registering voice data finely adjusted by the fine adjustment means as voice data in a voice dictionary.