JPS6059471A - Electronic translating machine - Google Patents

Abstract

PURPOSE:To separate the areas of recognition between characters and the voice input by reproducing the sounds supplied to a tape recorder through a microphone to perform the stop and reproduction every character for speech recognition to a voice input type word processor and storing successively the characters to a memory part. CONSTITUTION:The voice is recorded discretely to a tape 9a of a cassette tape recorder 9 through a microphone 1. Then the tape 9a is loaded to a recorder 10 and reproduced. When a speech recognizing circuit 2a detects a voice and then a blank area, a control circuit 11 stops the reproduction of the recorder 10. The character signal recognized by the circuit 2a is written to a memory 5 via a CPU4. When the recognizing action is through, the circuit 2a applies a signal to the recorder 10 to restart the reproduction for recognition and storage of the next voice. The character signal reproduced completely and written to the memory 5 is displayed to a CRT7 or hard-copied by a typewriter 6. If the displayed character has some error, a key input part 8 corrects it.

Description

[Detailed description of the invention] The present invention relates to an electronic translation device.

In addition, in this specification, translation is understood in a broad sense, and (1)
(2) converting the language of one country into the language of another country; (2) mutually converting an abbreviation and the original language, or mutually converting a language and its explanation, even if the language is within the same country; (3) converting characters and Includes mutual conversion with speech, (4) computation, (5) other information processing, pattern recognition, etc.

The source language refers to the language, numbers, formulas, etc. before translation, and the translated language refers to the language, numbers, formulas, etc. after translation. for example,
In a voice input type word processor (hereinafter referred to as WP), the voice ``I am Yamada.'' is the original language, and the voice ``I am Yamada.''
I'm Yamada. ” is the translation language. Furthermore, in a Japanese-English electronic translation device, "I AM YAMADA" is the original language, and "I AM YAMADA" is the translated language.

As an example of a conventional electronic translation device, a voice input type word processor (WP), which has already been put into practical use, will be explained with reference to FIG.

FIG. 1 is a block diagram of a conventional voice input type WP.

Microphone (hereinafter also referred to as MIC) captures discretely uttered sounds.
1, the voice is recognized by a voice recognition circuit (hereinafter referred to as SA) 2 using the DP matching method, and converted into a character signal. , FD), RAM, etc., a word processor (WP) 3 consisting of an electronic typewriter (hereinafter referred to as TW) 6, and a cathode ray tube (hereinafter referred to as CRT) 7, and output to CRT 7 or TW 6. .

Note that 8 is a key input section used when, for example, the SA 2 erroneously recognizes the voice applied from the microphone 1.

By doing so, the voice applied to the microphone 1 can be transcribed into text.

However, in such a voice input type WP, the microphone 1 is connected to the WP 3 by wire or wirelessly via the SA2, so the speaker is always present at the microphone 1 and speaks while looking at the manuscript or without the manuscript. The disadvantage was that it required vocalization.

That is, since the utterance operation and the recognition operation were performed at the same time, there was a drawback in that the speaker had to be at the WP installation location at all times during input.

The present invention eliminates these drawbacks and provides an electronic translation device that can break the inseparable relationship between the speaker and the WP.

An embodiment of the present invention will be described below with reference to FIGS. 2 and 3.

FIG. 2 is a block diagram of a voice input type WP which is an embodiment of the present invention, and FIG. 3 is a timing chart of input/output signals of each part of the WP.

In the figure, 1 is a microphone connected to a cassette tape recorder (hereinafter referred to as TP) 9, and 9a is a cassette containing a magnetic tape, which can be detachably attached to the TP 9.

Reference numeral 10 denotes a temporarily stopable cassette tape recorder (hereinafter referred to as ASTP), into which a cassette 9a on which an audio signal has been recorded by TP 9 is mounted, and the audio signal recorded on the cassette 9a is reproduced.

ASTP10 is connected to WP3 via SA2a. SA2a generates a pulse signal when a blank portion of the input audio signal is detected and when the recognition operation ends.

This pulse signal controls the control circuit (hereinafter referred to as CT).
11 controls the ASTP 10 to start or stop the playback operation. Note that WP3, like WP3 shown in Fig. 1, has CPU4, memory 5, KEY8, CRT7 and T
It is composed of W6 and the like.

Next, the operation of this device will be explained.

First, the voice ``I am Yamada'' was recorded in discrete utterances through the microphone 1 into the cassette 9a attached to the TP9.
Furthermore, remove cassette 9a from TP9, and
Let us consider the case where the cassette 9a is attached to the cassette 9a and the recorded portion is played back.

As shown in FIG. 3(A), the audio signal ``Watashiwayamadadesu'' is continuously recorded in discrete utterances in the cassette 9a. When reproducing the audio signal "wa", there is a blank time of about 0.2 seconds or more after "wa", so the SA2a detects this blank part and generates a pulse as shown in Figure 3(C).
As a result, CT11 becomes AST as shown in 3rd (F).
The playback operation of P10 is stopped.

Then, it recognizes that SA2a is a voice signal "wa" and writes a character signal "wa" into the memory 5 via the CPU 4 as shown in FIG. 3(D). When SA2a completes the recognition as described above, SA2a returns to FIG. 3 (E
) A pulse is applied to CT11 as shown in FIG.

As a result, the CT 11 puts the ASTP 10 into the reproduction state again, and reproduces the next audio signal "ta". The reproduced audio signal "ta" is applied to the SA 2a, converted into a character signal "ta", and then written into the memory 5 via the CPU 4. At this time, SA2a generates a pulse before and after recognition, and CT11 accordingly
0 is stopped and played again.

In this way, the audio recorded in cassette 9a one after another is
ASTP10 uses “wa” and “ta” for “watashiwayamadades”.
, "shi", "wa", "ya", "ma", "da", "de",
The audio signals are reproduced while being divided into "su", and after these divided audio signals are converted into character signals via the SA 2a, they are written into the memory 5 from the CPU 4.

Then, when it is played to the end by ASTP10, the CPU
4 displays all the character signals written in the memory 5 on the CRT 7 or in a preprogrammed order and arrangement.
Make a hard copy using W6 and display the text.

In this way, it is checked later whether the characters displayed on the CRT 7 or the TW 6 match the voice input from the microphone 1, and if there is an error, it is corrected using the KEY 8.

Furthermore, the above character display is automatically converted into kanji by a kana-kanji conversion function built into the CPU 4.

In this way, the speaker can record the words to be typed (written into sentences) on a tape recorder in advance, and then simply attaches the tape to the WP, and then the WP automatically records the words even when the speaker is not present. Therefore, the speaker's constraints on time and location can be relaxed, and other business processes can be carried out during this time.

Next, another embodiment of the present invention is shown in FIG. I will explain using

FIG. 4 is another embodiment of the present invention. FIG. 2 is a block diagram of an electronic translation device.

1 in the figure is a cassette tape recorder (ASTP) 10
2a is a voice recognition circuit (SA) using the DP matching method; 11 is an AS that follows the pulses generated from SA2a in response to the recognition operation of SA2a;
This is a control circuit (CT) for switching the TP 10 to a pause or playback state. ASTP10, SA2a, CT1
1 is similar to that of the previous embodiment, and its operation is also the same.

12 is a Japanese-English electronic translation circuit (hereinafter referred to as TI) for translating the character signal (Japanese) from SA2a into an English character signal.
), and 19 is a key input unit (KEY) for operating the TI 12 by key operation separately from the SA 2a.

13 is a memory (
14 is an electronic typewriter (TW) for hard copying the character signals from the TI 12, and 15 is a cathode ray tube (CRT) for projecting the character signals as characters and making them visible. be.

Further, 16 is a speech synthesis circuit (hereinafter referred to as SM) for converting the character signal from the TI12 into a voice signal, and 17 and 18 are respectively connected to the SM16.
tape recorder (TP) for recording audio signals from
) and a speaker (SP) for listening to the same audio signal
It is. In addition, KEY19, memory 13, TI12, SM
16, TW14, CRT15, TP17, or SP18 has already been put into practical use, so a detailed explanation of this part will be omitted here.

Next, the operation of this device will be explained.

First, record the sentence you want to translate, such as "I am Yamada", into ASTP10 via microphone 1. Next this AS
When TP10 is played back (the speaker may not be present at this time), the continuous voice from ASTP10, "watashi wa yama da desu" will be heard as "wa", "ta", "shi" in order to control ASTP10 by SA2a and CT11. ..., and the same operation as explained using FIG. ","Ma","
Character signals such as ``da'', ``de'', and ``su'' are applied to the TI 12 and stored in the memory 13.

Next, upon detecting this end, the TI 12 translates this character signal (Japanese) into a character signal "I AM YAMADA" (English) and sends it to the TW 14, CRT 15, and SM 16, respectively.

TW14 types out this character signal and also CRT1
5 displays the same characters. Further, the SM 16 converts this character signal into a voice signal "I am Yamada" and records it in the TP 17, and causes the SP 18 to utter it.

By doing this, the speaker only needs to record sentences, words, phrases, etc. that they want to translate into ASTP 10 through microphone 1, and then the translated sentences will be automatically converted into text or audio and recorded even if they are not present. You can keep it.

In this case, if the present inventor implements the technology already proposed in JP-A-57-17081 and JP-A-57-100561, it will be easy to automatically give translation instructions to the TI 12 for each sentence break. Automatic translation can be performed. In addition, it has already been published in Japanese Unexamined Patent Publication No.
If the techniques proposed in No. 4, No. 57-57375, No. 57-122497, and No. 58-54483 are applied to this embodiment, a more advanced electronic translation device can be easily realized.

Next, another embodiment of the present invention will be described using FIG.

FIG. 5 is a block diagram of an electronic device with voice recognition which is another embodiment of the present invention.

In the figure, 1 is a microphone, 21 is a TP for recording the audio from the microphone 1, and 22 is a floppy disk (hereinafter referred to as FD) device that can temporarily record the audio signal from the TP 21 and take it out instantly if necessary. SA2a and CT11 are the same as those in the previous example, and FD22, SA2a and CT1
1 to form a closed loop, CT11 controls FD22 based on the instruction pulse from SA2a, and efficiently controls SA2a.
The voice signal is reproduced from the FD 22 one section at a time.

By doing this, if the SA2a makes an incorrect recognition (including unrecognized), the CT11 controls the FD22 to play the same part again, makes the SA2a send out the same audio signal again, and prevents errors caused by noise etc. In addition, in this case, the first time is compared, and the audio signal to be output is divided into short parts and sent out. In other words, the first time the same part is played back, it is played word by word, for example, "Yamada," but the second time, it is played word by word, for example, "ya,""ma," and "da." It makes it easy to recognize.

In this way, the character signal voice-recognized and converted by the SA 2a is applied to an external device (eg, WP, TI, teletype, etc.) 30, and information processing is performed thereon.

In the above three embodiments, an example was shown in which the audio signal of the microphone 1 is temporarily recorded on a tape recorder, but the invention is not limited to this, and it is also possible to use a floppy disk, RAM, or other recording/playback device. good.

In addition, although we have shown an example of a device that starts and stops the tape as an ASTP, it is not limited to this, and in addition to starting and stopping, if the device cannot be recognized, it may be necessary to rewind the tape a little and play the same part again, or to play the same part again. It is also possible to use a device that reproduces portions at a slower reproduction speed. In the above embodiment, an example of an audio signal is shown, but a video signal from a VTR or the like may be used for pattern recognition or graphic recognition instead of audio recognition.

Further, although an example has been shown in which the input voice is uttered discretely, it is also possible to record the input voice continuously on the tape and divide the playback from ASTP into discrete utterances. If you do this, SA
Even if 2a is a discrete utterance type SA, the speaker is not required to be in front of WP3, and once the sentence is recorded at normal speed, it can be transcribed into text even if the speaker is not present. Also, after the final check, if there is an error, the key is
You only need to correct that part in step 8.

Furthermore, instead of the discrete voice type SA, a continuous voice type SA may be used as the SA.

As described above, according to the present invention, since speech is once recorded on the recording means and speech recognition is performed while playing it back, there is no reduction in the recognition rate due to noise surrounding the speech recognition device, and the speaker can listen to the speech at any time. When a sentence is conceived, it can be recorded on tape, for example, and there is no need to restrain the speaker at the location where the device is installed during recognition or transcription.

It also becomes easier for the speaker to transcribe text over the phone. Since it is not possible to determine the availability of a WP over the phone, you have to wait for the WP to become available or make multiple phone calls to instruct the WP to write text.

Using the present invention, a speaker can use ASTP or T
All you have to do is record the sentence you want to transcribe into P, regardless of whether WP is active or empty, and WP will begin transcribing the speaker's sentences as soon as it becomes available. Therefore, one WP can be used effectively.

[Brief explanation of drawings]

Figure 1 is a block diagram of a conventional voice input type word processor, Figure 2 is a block diagram of a voice input type word processor which is an embodiment of the present invention, and Figure 3 shows the input and output signals of each part of the processor. The timing charts, FIGS. 4 and 5 are block diagrams showing other embodiments of the present invention. 1------Microphone, 2, 2a---Speech recognition circuit (SA), 3------Word processor (WP)
, 4------- Central processing circuit (CPU), 5,
13-----Memory, 6, 14--Electronic typewriter (TW), 7, 15-----Cathode ray tube (CRT), 8
, 19---Key input section (KEY), 9, 17, 21
--- Tape recorder (TP), 9a --- Cassette, 10 --- Tape recorder (ASTP), 1
1---Control circuit (CT), 12---Electronic translation circuit (TI), 16------
Speech synthesis circuit (SM), 18------- Speaker (S
P), 22-----Floppy disk (FD), 30--
---External equipment. Patent applicant's name Noriko Ikegami

Claims (3)

[Claims] (1) A storage unit for storing voice signals; a voice recognition unit for recognizing the voice signals from the storage unit and converting them into character signals; and a control section for controlling the storage section to cause the speech recognition section to send a speech signal from the speech recognition section. (2) The electronic translation device according to claim 1, wherein the speech recognition section includes a character display section for displaying the character signal as characters. (3) a storage unit for storing voice signals; a voice recognition unit for recognizing the voice signals from the storage unit and converting them into character signals; a control unit that controls the storage unit to cause the unit to send a voice signal to the voice recognition unit; an electronic translation unit that translates a character signal of the voice recognition unit into a character signal of a translation language; an electronic translation device comprising: a character display section for displaying a character signal of the translated language; or a translated language output section having a speech synthesis section for converting the text signal of the translated language into an audio signal.