JPS636355B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a kanji input device for converting speech into kanji codes and inputting them into a computer phototypesetting system.

Conventionally, when processing kanji using a computer, an operator inputs kanji data using a kanji keyboard or a kanji tablet using his or her hands or fingers. Therefore, inputting kanji data requires a great deal of effort, requires technical skill, and is physically tiring. Therefore, an object of the present invention is to provide a kanji input device for printing that does not have these drawbacks.

This invention will be explained below.

The present invention relates to a printing kanji input device for converting speech into kanji codes and inputting them into a computerized phototypesetting system.As shown in FIG. The device 10 stores phoneme parameters of a predetermined phoneme along with a phoneme code. The phoneme data storage device 20 compares the feature parameter CP from the parameter extraction device 10 with the phoneme parameter SP from the phoneme data storage device 20, and generates a phoneme code corresponding to the phoneme parameter having the highest similarity and having a predetermined degree of similarity or higher. Audio determination device 3 that outputs SC
0, the monitoring device 200 that monitors the opening and closing of the lips 101 of the speaker 100, and the speech determination device 30 output a phoneme code SC that is difficult to distinguish, and the opening/closing signal NS from the lip opening/closing monitoring device 200 determines that the phoneme code is correct. A gate device 40 that outputs a phoneme code as a conversion phoneme code TSC, a Japanese language dictionary storage device 50 that stores descriptive word codes corresponding to various headwords as a Japanese language dictionary, and a conversion output from the gate device 40. Phoneme code TSC and Japanese dictionary storage device 5
0 is compared with the entry word, and the descriptive word data DW of the matching entry word is called and the font memory 2 is
The kana-kanji conversion device 60 outputs to the display device 3 according to the font information FI from A selection designation device 70 for inputting desired constituent characters when a corresponding word does not exist, and a kanji code CH output from the kana-kanji conversion device 60 according to the selection designation of this selection designation device are stored and computerized together with the allocation information. A kanji code storage device 80 for transmission to the phototypesetting system 4 is provided. put it here,
As shown in FIG. 2, the parameter extraction device 10
A sampling circuit 11 samples the audio signal AS from the microphone 1 at predetermined intervals (for example, 10 ms) according to the sampling signal SM, and the sampling data SD sampled by the sampling circuit 11 is passed through eight channels of bandpass filters 12F1 to 12F8. Filter 1 that waves with
2 and this filter 12 (12F1 to 12F8)
The quantization circuit 13 quantizes each output into audio quantum signals (CH1S to CH8S; 2 bits each) as characteristic parameters CP. The passband widths of the bandpass filters 12F1 to 12F8 constituting the filter 12 are as shown in FIG. 3, and are designed to equally divide the audio frequency of 200 to 5000 Hz logarithmically.

The lip opening/closing monitoring device 200 also includes a TV camera 201 for displaying the vicinity of the lips 101 of the speaker 100.
, an AD converter 202 that converts the video signal from this TV camera 201 into a digital signal, and this
It is comprised of a frame memory 203 that stores digital signals from the AD converter 202 frame by frame, and an open/close determination circuit 204 that determines whether the lips 101 are open or closed based on this stored information.

As shown in FIG. 4, the phoneme data storage device 20 stores 5 types (A~O) of vowel phoneme parameters (channels CH1~CH8) and 15 types (B~Z) of vowel phoneme parameters.
Consonant phoneme parameters (channel CH1~
CH8) are stored together with phoneme codes each expressed as a four-digit hexadecimal number according to JIS code. The Japanese language dictionary storage device 50 also stores entry words arranged in alphabetical order as shown in FIG.
This Japanese language dictionary storage device 50 has a function as a Japanese language dictionary or a Japanese language dictionary for daily use.

In such a configuration, a voice inputter (who converts the voice into a kanji code CHA and inputs it into the computer phototypesetting system 4)
When an operator (such as an operator inputting data into a computer) stands in front of the TV camera 201 and reads a predetermined sentence in front of the microphone 1, the audio signal AS from the microphone 1 is transmitted to the parameter extraction device 10. The signal is input to the sampling circuit 11 in the internal circuit and sampled. In other words, if the sentence is ``Cherry blossoms are flowers,'' it would be ``SAKURA HA TEN HANA.''
DESU MARU”. Here, the audio signal AS
For example, in the case of "E" (E), it becomes as shown in FIG. are respectively input to bandpass filters 12F1 to 12F8. Therefore, these band pass filters 12F1 to 12
F8 passes the sampling data SD for each band width shown in FIG.
, and a four-stage defect (00
~11) 2-bit audio quantum signal CH1S~CH8
Convert to S. Here, each of the 2-bit 8-channel audio quantum signals CH1S to CH8S
For example, as shown in FIG. 7, the feature parameter CP consisting of sampling times SAMP1, SAMP
2, . . . and input into the speech determination device 30. However, the voice determination device 30
The phoneme parameters of the worker who uses the kanji input device are stored, and for example, as shown in Fig. 4, the phoneme parameters and phoneme codes corresponding to vowels (A to O) and consonants (B to Z) are stored. It is stored for use by that worker. Then, the stored phoneme parameters CH1 to CH8 and the feature parameter CP (speech quantum signal
CH1S to CH8S) are compared for each sampling to determine the degree of similarity. In this case, the error range of ±1 out of 4 levels for each channel's data is considered to be the same, and the number of channels that are the same is counted, and the counted value is, for example, "6" (6 out of 8 channels). When the channels match), the phoneme code corresponding to that phoneme parameter is output. That is, for example, if the phoneme parameter Q is "00", "10", "11" from channel 1,
In the case of "01", "00", "10", "10", "01", "00",
Feature parameter CP is "01" from channel 1,
"10", "11", "10", "00", "10", "01", "
00", channels 1 and 4 have an error of ±1, and everything else matches (the count value at this time is "8")
Therefore, a phoneme code corresponding to the above phoneme parameter Q is output. Also, for the above phoneme parameter Q, the feature parameter CP is from channel 1 to "10", "01", "11", "11", "11", "00".
”,
In the case of "01" and "11", the number of matching channels is "2", so no phoneme code is output.
Such a comparison operation is performed for each phoneme parameter, and a phoneme code corresponding to the phoneme parameter with the highest number of matches among them is output.

Then, only when such a phoneme code continues for a predetermined time (for example, 30ms), that is, when the same phoneme code continues three or more times for sampling times SAM1, SAM2, ..., the phoneme code is summarized as a discriminated phoneme code SC. and output it. Therefore, if a comparison result as shown in FIG. 8A is obtained from the comparison between the feature parameter CP and the phoneme parameter SP, the phoneme code SC to be outputted will be a phoneme code string as shown in FIG.
Only when it continues more than once, it is output in a form as shown in C of the same figure. The phoneme code SC output in this manner is input to the gate device 40.

On the other hand, when an operator or the like inputs voice, the lips 10
1 is projected onto the TV camera 201, and its opening/closing status is monitored.The operation will be explained with reference to the flowchart of FIG.

First, the lips 101 of the speaker 100 such as an operator are displayed on the screen of the TV camera 201.
TV camera 201 is set. The image on the screen is converted into a digital signal by the AD converter 202, and the digital image signal and each frame are stored in the frame memory 203. Then, the open/close determination circuit 204 reads out and processes the information stored in the frame memory 203, determines whether the lips 101 are open or closed, and receives the corresponding open/close signal NS.
Output as . Furthermore, in processing the sound from the microphone 1, it is difficult to extract characteristic parameters for distinguishing between the nasal sounds "M" and "N" and the voiceless plosive sounds "P" and "T", making it difficult to distinguish between them. Therefore, it is known that the nasal sound "M" and the plosive sound "P" are produced when the lips 101 are closed, and the nasal sound "N" and the plosive sound "T" are produced when the lips 101 are open. , the "M" and "N" sounds are discriminated based on the opening/closing signal NS from the lip opening/closing monitoring device 200. That is, the gate device 40 is "M",
For phonemes other than the "N", "P", and "T" sounds, the output operation is performed by the speech determination device 30 as described above, and the nasal "M", "N" sounds or the plosive "P",
Lip opening/closing monitoring device 2 only when “T” sound is detected
Referring to the opening/closing signal NS from 00, the correct phoneme is output as the conversion phoneme code TSC.

The conversion phoneme code TSC from the gate device 40 obtained in this way is transferred to the kana-kanji conversion device 60.
is input. For example, if you input the sentence "Cherry blossoms are flowers," the conversion phoneme code TSC would be "SAKURA HA TEN HANA."
DESU MARU” and convert each of these phonemes into hexadecimal 4
The data is expressed in digit JIS code. In this way, the kana-kanji conversion device 60 sequentially transmits the input conversion phoneme codes TSC to the Japanese language dictionary storage device 50, checks whether or not there is an entry word that corresponds to the phoneme string, and if so, The descriptive word data DW corresponding to the headword is called each time. At the same time, the descriptor data DW is transferred to the font memory 2.
, the corresponding font information FI is called, and all of the descriptor word data DW is displayed on the display device 3 in accordance with this font information FI. In this case, the descriptive word data DW contains homophones (for example, "hashi", "edge", and "chopsticks" in Figure 5) or homophones (for example, "sakura", "sakura", and "cherry blossom" in Figure 5). When there is no word, the descriptor word data DW is directly output to the kanji code storage device 80 as the kanji code CH. Therefore, all of the descriptor data DW corresponding to the entry word, for example, the descriptor data DW for the entry word "SAKURA" are displayed on the display device 3.
The DWs “Sakura (2435, 242F, 2469),” “Sakura (2535, 252F, 2569),” and “Sakura (6115)” are displayed together with their serial numbers 1 to 3, so the operator can choose from among them. The word is selected by specifying the corresponding number using the selection specifying device 70. Thus, when a desired word is selected and specified by the selection and specification device 70, only the selected word is re-displayed on the display device 3, and its Kanji code CH is output and stored in the Kanji code storage device 80. Such a selection designation operation by the selection designation device 70 is performed sequentially according to the input conversion phoneme code TSC, and the selected kanji code CH
is stored in the kanji code storage device 80. In addition,
In addition to selecting and specifying the descriptor data DW, the selection specifying device 70 also selects and specifies quarters (a space for one character), carriage return (forcibly wraps to the beginning of the next line), ruby (furikana), and sidelines (ruled lines and Function commands such as line break (move to the next column), page break (move to the next page), etc., as well as symbols (for example, ",", (, ), ;) and pictures (for example, ■, 〓 ,〓,〓)
We also specify the terms of . Furthermore, even if there is no entry word registered in the Japanese dictionary storage device 50, or if the entry word is registered but has different pronunciations, the selection and designation device 70 can be used to input and designate the desired word. Become.

Kana-kanji conversion device 6 output in this way
The kanji code CH starting from 0 is temporarily stored in the kanji code storage device 80, and this stored data is output as the kanji code CHA of the computer phototypesetting system 4.
In this case, the computer phototypesetting system 4 includes the position and size of headings, the structure of columns and columns for one page, the number of characters in one line,
Layout information that specifies the layout of font size, type, etc.
LY is also entered at the same time.

As described above, according to the present invention, voice recognition is performed using the characteristic parameters of the voice signal from the microphone, and is also performed using the opening/closing signals from the speaker's lip opening/closing monitoring device, so that extremely accurate voice can be obtained. - Kanji code conversion can be realized.

The above explanation describes an example in which sentences are input in the so-called wakachi reading.
It is also possible to input a chi symbol (corresponding to the blank space in the waka chi reading). In addition, in the above embodiment, the voice characteristic parameters are obtained using an 8-channel bandpass filter, etc., but this is 16 channels.
It is possible to have 20 channels, and the number of bits in each channel of the audio quantum signal can also be arbitrary (for example, 8 bits).

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an embodiment of this invention, FIG. 2 is a block diagram showing an example of the configuration of a parameter extraction device, and FIG. 3 is an example of the bandwidth of a filter (8 channels) used in this invention. 4 is a diagram showing the state of data stored in the phoneme data storage device, FIG. 5 is a diagram showing an example of the storage state of the Japanese language dictionary storage device, and FIG. 6 is a diagram showing an example of the audio signal. 7A and 7B are diagrams showing an example of feature parameters that are output from the parameter extraction device, and FIG. 8A
-C are diagrams showing how the phoneme code is output from the speech determination device, and FIG. 9 is a flowchart showing the operation up to the output of the phoneme code of the present invention. 1...Microphone, 2...Font memory, 3...
Display device, 4... Computer phototypesetting system, 10... Parameter extraction device, 11... Sampling circuit,
12... Filter, 13... Quantization circuit, 20...
...Phoneme data storage device, 30...Speech determination device,
40...Gate device, 50...Japanese language dictionary storage device, 60...Kana-kanji conversion device, 70...Selection specification device, 80...Kanji code storage device, 100...
...Speaker (operator, etc.), 101...Lip, 200
...Lip opening/closing monitoring device, 201...TV camera, 2
02...AD converter, 203...Frame memory, 204...Opening/closing judgment circuit, CP...Characteristic parameter, SP...Phoneme parameter, PS...Paratograph signal, SC...Phoneme code, TSC...Conversion phoneme Code, DW...descriptive word data, FI...font information, AS...audio signal, SM...sampling signal, SD...sampling data, PD...
...paratograph data, LY...allocation information.

Claims (1)

[Scope of Claims] 1 (a) A parameter extraction device that extracts feature parameters of input speech; (b) A phoneme data storage device that stores phoneme parameters of a predetermined phoneme along with a phoneme code; (c) (d) monitoring the opening and closing of the speaker's lips; a lip opening/closing monitoring device; (e) a gate device that outputs a correct phoneme code based on an opening/closing signal from the lip opening/closing monitoring device when a phoneme code that is difficult to distinguish is output from the speech determining device; (f) (g) a phoneme code for conversion outputted from the gate device and the entry word of the Japanese dictionary storage device; (h) a kana-kanji conversion device that compares the descriptive word data of the corresponding headword and outputs it to a display device according to the font information from the font memory; a selection and designation device for selecting and designating one of the words and inputting a function, a punctuation, and a desired constituent character in the case where the corresponding word does not exist; A kanji input device for printing, comprising: a kanji code storage device for storing kanji codes output from a conversion device and transmitting the kanji codes together with layout information to a computer phototypesetting system.