JPH0558555B2 - - Google Patents

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention relates to a Chinese speech recognition device, and in particular,
This invention relates to a Chinese speech recognition device that is useful when searching for Chinese characters.

[Conventional technology]

Traditionally, a tablet method has been used to search for Chinese characters. In this tablet method of input, all Chinese characters are printed on the tablet, and codes are attached to the characters, the characters on the tablet are detected manually, and the kanji codes are input into the information processing system by pressing a pen-touch or similar device. Ru.

On the other hand, in Japanese, devices that recognize pronunciation at the word level and search at the word level have already been put into practical use.

[Problem that the invention seeks to solve]

However, until now, there has been no device that recognizes the Chinese characters that make up Chinese words by voice. Therefore, when inputting Chinese characters, it is necessary to detect one character from among all the Chinese characters on the tablet, resulting in a disadvantage that a large amount of input time is required.

[Means for solving problems]

The Chinese speech recognition device of the present invention includes a speech input means for converting speech into an electrical signal and performing analog-to-digital conversion, and a speech encoding control means for encoding the speech signal converted from analog to digital by the speech input means. a storage means for storing the encoded voice data encoded by the voice encoding control means; a phonetic symbol for Chinese, a tone, and a voice code that corresponds to the voice pronounced by recognizing these phonetic symbols and tone; a standard pattern file means for storing encoded data as a standard pattern in advance; a comparing means for comparing the audio encoded data as a standard pattern from the standard pattern file means and the audio encoded data from the storage means; When the comparison result shows a match,
and output means for outputting phonetic symbols and tone tones corresponding to the matched audio encoded data from the standard pattern file means as recognition results.

〔Example〕

Next, the present invention will be explained in detail with reference to the drawings.

Referring to FIG. 1, the system to which the present invention is applied includes a Chinese speech recognition device 1 which is an embodiment of the present invention, an information processing system 7, and a Chinese language speech recognition device 1 that is an embodiment of the present invention, an information processing system 7, and a Chinese language speech recognition device 1 that is an embodiment of the present invention. A kanji character pattern file 8 is provided for storing kanji/numeric character patterns.

One embodiment of the present invention includes a voice input section 2 for inputting voice, a keyboard 4 for inputting data,
CRT3 for displaying at least one of phonetic symbols, tones, Chinese characters, and input data;
It is comprised of a standard pattern file 6 for storing audio encoded data, logical addresses, phonetic symbols and tones in the order of the Mandarin phonetic table, and a control section 5 that controls these components.

Next, the operation of the Chinese speech recognition device 1 which inputs and recognizes this speech will be explained in detail with reference to FIGS. 1, 2, and 3. As a prerequisite for speech recognition, it is necessary to create a standard pattern file 6. This standard pattern file 6 is created as follows. Referring to FIG. 2, for example, when the sound "wo" is pronounced by combining the phonetic symbol "wo" and the tone "3rd tone (〓)", the keyboard 4 is used to display the phonetic symbol data "wo" and the tone symbol data. “3” is pressed. The data “wo” and “3” pressed in this way are given to the file control unit 57 via the pronunciation/tone symbol data section 582 of the keyboard control section 58, and are sent to the file control section 57 via the pronunciation/tone symbol data section 582, and are stored as standard data by the pronunciation/tone symbol writing section 572. It is stored at a predetermined address in the pattern file 6, for example address _n7 shown in FIG. At this time, the data "wo" and "3" are given to the file control unit 57, and a signal indicating that file creation processing is being performed is sent to the search/file control unit 57.
It is given to the file creation detection section 62.

The sound “wo” pronounced in this condition is
The audio input unit 2 converts the signal into an electrical signal, amplifies it, and converts it from analog to digital through a low-pass filter. An example of this is U.S. Pat. No. 4,239,936 No. 3.
Reference may be made to the figures and the description in column 4. The digital audio electrical signal converted in this manner is digitally encoded in the audio encoding control section 2 using, for example, an adaptive differential pulse code modulation (ADPCM) method. The audio encoded data digitally encoded in this manner is stored in the storage section 61 and then read out and provided to the search/file creation detection section 62. Since the search/file creation detection unit 62 detects that file creation processing is being performed using the above-mentioned signal, it supplies the given audio encoded data to the audio encoded data writing unit 571 of the file control unit 57 . This audio encoded data writing section 571 is
The audio encoded data is stored at a predetermined address in the standard pattern file 6, for example address _n7 shown in FIG. In this way, "A" shown in Figure 3,
"Japan", "Zai", "hon", "chu", "〓", "country", "
Sentence",
A set of audio encoded data, pronunciation symbols, and tones for each of "Gaku" and "Yoshi" is stored at a predetermined address, and a standard pattern file 6 is created. Next, speech recognition processing according to an embodiment of the present invention will be described in detail. First, the uttered voice "wo" is converted into a digital signal by the voice input section 2, digitally encoded by the voice encoding control section 2, and stored in the storage section 61 as voice encoded data. At this time, since no data is provided from the keyboard 4, the audio encoded data read from the storage section 61 is detected as a search process by the search/file creation detection section 62, and is provided to the comparison control section 51. On the other hand, audio encoded data is sequentially read out from the standard pattern file 6 by the audio encoded data reading section 573 and provided to the comparison control section 51. The comparison control section 51 sequentially compares the data provided from the search/file creation detection section 62 and the data sequentially provided from the audio encoded data reading section 573. The match determination unit 52 determines whether the data matches or does not match based on the comparison result data. Unless it is determined that there is a match, the comparison end detection unit 5
The comparison operation is repeated by the comparison control unit 51 until the data is determined to be the final data of the standard pattern file 6 at step 3. As a result, if it continues to be determined that the data do not match up to the final data, the comparison end detection section 53 generates an end signal. In response to this end signal, the error display control section 55 sends an error message reading instruction to the error message reading section 575. An error message read instruction is sent to the error message read section 575. In response to this instruction, the error message reading section 575 reads out the error message from the standard pattern file 6 and sends it to the error message display section 562. The error display control section 55 responds to the end signal to display the error message display section 5.
A display instruction is output to 62. In response to this display instruction, the error message display section 562 displays the error message provided from the error message reading section 575 on the CRT 3. This error message indicates that the Chinese pronunciation is bad and is not stored in the standard pattern file 6.
Those searching by pronunciation will try again to use the correct Chinese pronunciation.

On the other hand, if the match determination unit 52 determines that there is a match, the phonetic symbol generation instruction unit 54 sends the phonetic symbol generation instruction unit 54 to the file control unit
A reading instruction is issued to the pronunciation/tone symbol reading unit 574 of No. 7. In response to this read instruction, the pronunciation/tone symbol reading section 574 reads out the pronunciation symbol "wo" and the tone "3" from the standard pattern file 6, and sends them to the information processing system 7 via the transmission section 591 of the communication control section 59. Send to. At the same time, the phonetic symbol generation instruction section 54 gives a display instruction to the pronunciation/tone symbol display section 561 of the CRT control section 56, and this pronunciation/tone symbol display section 561 outputs the phonetic symbol "wo" from the pronunciation/tone symbol reading section 574. ” and tone “3” are displayed on the CRT3.

Next, the phonetic symbol “wo” which is an application example of the present invention.
The configurations and operations of the information processing system 7 and the Kanji character pattern file 8 given the tone "3" will be explained in detail with reference to FIGS. 1, 4, and 5. Referring to FIGS. 1 and 4, the pronunciation symbol “wo” and tone “3” given from the control unit 5 of the Chinese speech recognition device 1 are transmitted to the storage unit via the communication control unit 71 of the information processing system 7. 72. The pronunciation/tone comparison unit 73 compares the phonetic symbol “wo” and tone “3” read from the storage unit 72 with the phonetic symbol “A” and tone “1” shown in FIG. 5 of the kanji character pattern file 8. be compared. Since the comparison result is a mismatch, the end detection unit 74 is notified of the mismatch. The end detection section 74 is a pronunciation/tone comparison section 73,
All phonetic symbols of Kanji character pattern file 8/
It is determined whether the comparison between the tone and the pronunciation symbol/tone given from the storage section 72 has been completed, and if the comparison has not been completed, the pronunciation/tone comparison section 73 is instructed to repeat the comparison operation sequentially until the comparison is completed. The pronunciation/tone comparison unit 73 compares the phonetic symbol “wo” and tone “3” read from the storage unit 72 with the phonetic symbol “wo” and tone “3” shown in FIG. 5 of the kanji character pattern file 8. are compared. The comparison result shows a match, so the phonetic symbol “wo” shown in Figure 5
The kanji character pattern “1. Ga” in the same group as tone “3” is stored in the storage unit 72. This kanji character pattern "1. w" is sent from the storage section 72 to the control section 5 via the communication control section 71.

The pronunciation/tone comparing unit 73 compares all the phonetic symbols/tones from the kanji character pattern file 8, and as a result, no match is found, and the end detecting unit 74
When the end is detected, the error message generating section 75 generates an error message. This error message is given to the control section 5 via the communication control section 71. Next, the process of selecting and displaying the Kanji character patterns and error messages read out from the information processing system 7 will be described in detail with reference to FIGS. 1, 2, and 5.

First, the display processing of Kanji character patterns will be explained below. Referring to FIGS. 1 and 2, the kanji character pattern "1. が" given from the information processing system 7 is transmitted to the receiving section 592 of the communication control section 59 and the kanji character pattern display section 5 of the CRT control section 56.
63 to be displayed on the CRT3. At this time, since the kanji character pattern is only one character "1. w", there is no need to perform selection processing. In addition to this kanji character pattern "1. w", the CRT screen also displays the phonetic symbol "wo" which is read out from the standard pattern file 6 by the pronunciation/tone symbol reading section 574 and displayed on the CRT 3 via the pronunciation/tone symbol display section 561. ” and tone “3” are displayed. Therefore, in addition to "wo" and "3", "1. w" is displayed on the CRT screen.
Next, an example of selection processing required for displaying a Kanji character pattern will be explained.

Since the search is “Japanese Japanese Studies〓Chinese”,
Next to "wa", phonetic symbol data "ZAI" and tone symbol data "4" are input from the keyboard 4, and "zai" is pronounced and input by the voice input section 2. Thereafter, the phonetic symbol "ZAI" and the tone "3" are read out from the logical address _n2 of the standard pattern file 6 in the same manner as described above, and are displayed on the CRT 3 and sent to the information processing system 7. In the information processing system 7, the kanji character pattern ``1. re, 2. zai, 3.〓'' that matches the phonetic symbol ``ZAI'' and the tone ``3'' is read out from the kanji character pattern file 7 and sent to the control unit 5. be done. The control unit 5 sends “1. re, 2. present, 3.〓” to the CRT 3 via the receiving unit 592 and the kanji character pattern display unit 563
is displayed. The searcher looks at this CRT3 screen,
Enter the numerical data "3" from the keyboard 4.
This numerical data “3” is given to the Kanji character pattern selection unit 564 of the CRT control unit 56 via the numerical data unit 581 of the keyboard control unit 58. The kanji character pattern selection section 564 selects the pattern "1." displayed by the kanji character pattern display section 563.
Kanji character pattern display section 563 instructs selection to display only "3.〓" among "re, 2.zai, 3.〓".
give to In response to this instruction, the kanji character pattern display section 563 selects only “3.〓” and displays the CRT3.
to be displayed.

Next, the processing by the control section 5 of an error message generated when the phonetic symbol and tone given from the control section 5 to the information processing system 7 are not in the Kanji character pattern file 8 will be explained. Referring to FIG. 2, the error message generated from the information processing system 7 of FIG. .
At this time, the CRT 3 includes a pronunciation/tone symbol reading section 57.
4 is read from the standard pattern file 6,
The pronunciation symbols and tone sent via the pronunciation/tone symbol display section 561 are displayed. Therefore, the CRT 3 displays the phonetic symbols, tones, and error messages in a composite manner. As a result, the searcher recognizes that although the Chinese pronunciation is good, there is no pronounced Chinese Kanji character pattern in the Kanji character pattern file 8. By using the above-described method, the remaining "Nihon Kokugaku (Chinese studies) in Japan" out of "I am in Japan (Chinese studies in Japanese) (Chinese)" is also searched word by word as follows. That is, "ri" is input from the voice input unit 2, and the phonetic symbol "RI" and tone "4" are input from the keyboard 4.
In response to this input, the control section 5 reads out the phonetic symbol "RI" and the tone "4" at address _n1 of the standard pattern file 6, displays them on the CRT 3, and sends them to the information processing system 7. The information processing system 7 reads the kanji character pattern "1. 日" from the kanji character pattern file 8 based on the phonetic symbol "RI" and the tone "4", and the control unit 5
Display on CRT3 via .

Next, "ben" is input from the voice input section 2, and the phonetic symbol "BEN" and tone "3" are input from the keyboard 4. In response to this input, the control unit 5 selects the address n ₃ of the standard pattern file 6.
The phonetic symbol "BEN" and tone "3" are read out, displayed on the CRT 3, and sent to the information processing system 7. The information processing system 7 reads out the kanji character pattern "1.hon, 2.〓, 3.〓" from the kanji character pattern file 8 based on the phonetic symbol "BEN" and the tone "4", and reads it through the control unit 5. and display on CRT3. When the searcher types the numerical data "1" on the keyboard 4, only the kanji character pattern "1.hon" is selected.
Displayed on CRT3, other kanji patterns “2.〓, 3.
〓” will be deleted from the CRT3 screen.

Then, "guo" is input from the voice input section 2, and the phonetic symbol "GUO" is input from the keyboard 4.
and tone "2" are input. In response to this input, the control unit 5 outputs the phonetic symbol "GUO" and tone "2" at the address _n6 of the standard pattern file 6.
is read out, displayed on the CRT 3, and sent to the information processing system 7. The information processing system 7 generates a kanji character pattern from the kanji character pattern file 8 based on the phonetic symbol ``GUO'' and tone ``2''. .〓、
7. 〓” is read out and displayed on the CRT 3 via the control unit 5. When the searcher types the numerical data "1" on the keyboard 4, only the kanji character pattern "1. 国" is selected and displayed on the CRT 3, and other kanji patterns "2. 〓, 3. grab, 4. 〓、5.〓、6.〓、
7.〓” will be deleted from the CRT3 screen.

After that, "xue" is input from the voice input section 2, and the phonetic symbol "XUE" is input from the keyboard 4.
and tone "2" are input. In response to this input, the control unit 5 outputs the phonetic symbol "XUE" and tone "2" at address _n9 of the standard pattern file 6.
is read out, displayed on the CRT 3, and sent to the information processing system 7. The information processing system 7 generates the kanji character pattern ``1.ana, 2.〓, 3.〓, 4.gaku, 5.〓, 6 from the kanji character pattern file 8 based on the phonetic symbol ``XUE'' and tone ``2''. .〓、7.
〓” is read out and displayed on the CRT 3 via the control unit 5. When the searcher types the numeric data "4" on the keyboard 4, only the kanji character pattern "4. 学" is selected and displayed on the CRT 3, and other kanji patterns "1.ana, 2.〓, 3. 〓、5.〓、6.〓、
7.〓” will be deleted from the CRT3 screen.

Next, "xi" is input from the voice input section 2, and the phonetic symbol "XI" and tone "2" are input from the keyboard 4. In response to this input, the control unit 5 reads out the phonetic symbol "XI" and tone "2" at address _n5 of the standard pattern file 6,
Display on CRT3 and information processing system 7
Send to. The information processing system 7 generates Kanji character patterns "1.〓, 2.
〓, 3.〓, 4. seat, 5.〓, 6.〓, 7.〓, 8.〓, 9.〓”
is read out and displayed on the CRT 3 via the control unit 5.
When the searcher types the numerical data "1" on the keyboard 4, only the kanji character pattern "1." is selected and displayed on the CRT 3, and other kanji patterns "2.〓, 3.〓, 4." are selected. , 5.〓, 6.〓, 7.〓, 8.
〓, 9.〓” are deleted from the CRT3 screen.

Thereafter, "zhong" is input from the voice input section 2, and the phonetic symbol "ZHONG" and tone "1" are input from the keyboard 4. In response to this input, the control section 5 reads out the phonetic symbol "ZHONG" and the tone "1" at address _n4 of the standard pattern file 6, displays them on the CRT 3, and sends them to the information processing system 7. The information processing system 7 creates a kanji character pattern file 8 based on the phonetic symbol "ZHONG" and tone "1".
Kanji character pattern from "1. middle, 2. chu, 3. 〓, 4. 〓,
5. 辷, 6. 松, 7. 端, 8.〓” are read out and displayed on the CRT 3 via the control unit 5. When the searcher types the numerical data "1" on the keyboard 4, only the kanji character pattern "1. 中" is selected.
Displayed on the CRT3, other kanji patterns "2. Chu, 3.
〓, 4. 〓, 5. 衷, 6. 松, 7. 〓, 8. 〓 are deleted from the CRT3 screen.

Finally, “wen” is input from the voice input section 2, and the phonetic symbol “WEN” is input from the keyboard 4.
and tone "2" are input. In response to this input, the control unit 5 outputs the phonetic symbol "WEN" and tone "2" at address _n8 of the standard pattern file 6.
is read out, displayed on the CRT 3, and sent to the information processing system 7. The information processing system 7 generates a kanji character pattern from the kanji character pattern file 8 based on the phonetic symbol ``WEN'' and tone ``2''. . Listen, 7.〓”
is read out and displayed on the CRT 3 via the control unit 5.
When the searcher types the numerical data “1” on the keyboard 4, the kanji character pattern “1. sentence” is generated.
is selected and displayed on the CRT3, and other kanji patterns "2.〓, 3.〓, 4.mosquito, 5.〓, 6.mon, 7.〓"
will be deleted from the CRT3 screen.

〔Effect of the invention〕

According to the speech recognition device of the present invention, if you speak in the standard language used in China, the speech is automatically input in a form that can be processed, so you can search for Chinese characters/numbers quickly. This has the effect of allowing you to perform searches. Compared to the conventional tablet system, the use of the device of the present invention has the effect of achieving a speed increase of 2.5 to 6 times.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing a system applied to the present invention, FIG. 2 is a diagram showing the configuration of an embodiment of the present invention, and FIG. 3 is a diagram showing the storage format of the standard pattern file of the embodiment shown in FIG. The figures shown, and FIGS. 4 and 5 are diagrams for explaining an application example of the present invention. In FIGS. 1 to 5, 1...Chinese speech recognition device, 2...Speech input unit, 3...CRT,
4...Keyboard, 5...Control unit, 6...Standard pattern file, 7...Information processing system, 8...
... Kanji character pattern file, 51 ... Comparison control section, 52 ... Match judgment section, 53 ... Comparison end detection section, 54 ... Phonetic symbol generation instruction section, 55 ... Error display control section, 56 ... CRT control Department, 57...
File control section, 58...Keyboard control section, 5
9... Communication control unit, 60... Audio encoding control unit,
61...Storage unit, 62...Search/file creation detection unit, 71...Communication control unit, 72...Storage unit, 7
3...Pronunciation/tone comparison section, 74...End detection section,
and 75...error message generating section.

Claims (1)

[Scope of Claims] 1. Audio input means for converting audio into an electrical signal and performing analog-to-digital conversion; audio encoding control means for encoding the audio electrical signal converted from analog to digital by the audio input means; storage means for storing encoded voice data encoded by the voice encoding control means; storing a plurality of the encoded voice data and phonetic symbols and tones corresponding to the encoded voice data as standard patterns; standard pattern storage means; first processing means for determining whether or not the phonetic symbol and tone corresponding to the audio encoded data have been input; a second processing means for registering the encoded speech data read from the storage means in the standard pattern storage means together with the corresponding phonetic symbol and the tone; If the processing means determines that the phonetic symbol and the tone have not been input, the encoded voice data read from the storage means and the encoded voice data read from the standard pattern storage means. a control means provided with a third processing means for sequentially comparing the results of the comparison; and a control means comprising: a third processing means for sequentially comparing the pronunciation symbols and tones corresponding to the encoded speech data that match as a result of the comparison by the third processing means from the standard pattern storage means; A Chinese speech recognition device characterized by comprising an output means for reading and outputting.