JPS61231629A - Voice input device - Google Patents

Voice input device

Info

Publication number
JPS61231629A
JPS61231629A JP60072637A JP7263785A JPS61231629A JP S61231629 A JPS61231629 A JP S61231629A JP 60072637 A JP60072637 A JP 60072637A JP 7263785 A JP7263785 A JP 7263785A JP S61231629 A JPS61231629 A JP S61231629A
Authority
JP
Japan
Prior art keywords
voice
input
closest
recognition
data
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
JP60072637A
Other languages
Japanese (ja)
Inventor
Isamu Muto
勇 武藤
Kaneo Tsukui
津久井 金雄
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hitachi Ltd
Original Assignee
Hitachi Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Hitachi Ltd filed Critical Hitachi Ltd
Priority to JP60072637A priority Critical patent/JPS61231629A/en
Publication of JPS61231629A publication Critical patent/JPS61231629A/en
Pending legal-status Critical Current

Links

Abstract

PURPOSE:To decrease the repetition of mis-recognition by recognizing an input voice, responding to it with the closest voice pattern, comparing the next input voice and the preceding voice when the input is recognized as an error and responding to it with the next closest voice pattern when they are coincident. CONSTITUTION:A voice input/output device 1 of a radio equipment fixed station 2 compares and collated input data with a registered voice pattern data in advance and stores the closest voice pattern and the 2nd closest voice pattern in a storage secton 9. A recognition device 6 recognizes the received voice data in the device 1 and inputs the result to a control section 8. The control section 8 extracts the closest voice pattern stored (9) from the result of recognition, a voice synthesizer 7 synthesizes it and returns the reply voice. A speaker 5 of a radio mobile station 3 listens to the response and it is discriminated as mis-recognition, a new voice signal is sent to represent the notice. The device 6 of the fixed station 2 compares and collates the voice signal inputted newly with the preceding voice signal, and when they are coincident, the control section 8 extracts the next closest voice pattern from the storage section 9, synthesizes (7) them and sends the result to the mobile station 3.

Description

【発明の詳細な説明】 〔発明の利用分野〕 本発明は、人間が音声にてデータを入力する音声入出力
装置の改良に関する。
DETAILED DESCRIPTION OF THE INVENTION [Field of Application of the Invention] The present invention relates to an improvement of a voice input/output device in which a human inputs data by voice.

〔発明の背景〕[Background of the invention]

従来の音声入出力装置は、音声入力したデータが誤認識
された場合、話者は、そのデータを訂正し、再度音声入
力しなおすが、再入力した時、同じ誤認識が繰り返され
る場合がある。これでは、話者の方も戸惑ってしまい、
音声入力の効率も悪くなる。
With conventional voice input/output devices, if voice input data is misrecognized, the speaker corrects the data and re-inputs the voice, but when the data is input again, the same misrecognition may be repeated. . This confuses the speaker,
The efficiency of voice input also deteriorates.

い第1の音声パターンデーダと、次に近い第2の音声パ
ターンデータとを判断しておき、第1の音声パターンに
対応するアンサーバックに対して、操作者から、誤認識
である旨の返答があった場合、上記第2の音声パターン
に対応するアンサーバックを自動的に行ってみて、操作
者から、「OK」である旨の返答を待つことが考えられ
る。
The first voice pattern data that is closest to the first voice pattern data and the second voice pattern data that is the next closest are determined, and in response to the answer back corresponding to the first voice pattern, the operator responds to the effect that it was misrecognized. If there is, it may be possible to automatically perform an answerback corresponding to the second voice pattern and wait for an "OK" response from the operator.

しかしながら、音声認識装置は、その認識率を高めるこ
とが極めてむず゛かしく、不特定話者用としては1例え
ば、子種類以下の言葉についてのみ、所定の認識率を得
ることができ、特定話者用としく2) ても、できる限り、限定された言葉に絞る必要がある。
However, it is extremely difficult to increase the recognition rate of speech recognition devices; 2) However, it is necessary to use limited words as much as possible.

とのゆ馬な観点で見ると、音声認識装置油しては、なる
べく少ない音声パターン数の中から対応音声パターンを
探し出すように□構成することが望ましく、同一の装置
にあっても、手順の進行の過程によっても、捜し出すべ
き音声パターン群を絞り込んでゆくことが望ましい。
From this point of view, it is desirable to configure a speech recognition device so that it searches for a corresponding speech pattern from among as few speech patterns as possible, and even if the device is the same, the procedure may be different. It is desirable to narrow down the group of voice patterns to be searched for as the process progresses.

このような意味から言えば、前述した従来技術の場合、
アンサーバックされた第2の音声に対する操作者の確認
の音声など、いたずらに、認識すべき音声パターン群を
拡大することとなり、高い認識率を達成する上では逆効
果となる欠点をもっている。
In this sense, in the case of the prior art mentioned above,
This method unnecessarily expands the group of voice patterns to be recognized, such as the voice of the operator confirming the second voice that has been answered back, and has the drawback of having the opposite effect in achieving a high recognition rate.

〔発明の目的〕[Purpose of the invention]

本発明の目的は、音声入力の際、簡単にして、誤認識を
繰り返すことの少ない音声入力装置を提供することにあ
る。
SUMMARY OF THE INVENTION An object of the present invention is to provide a voice input device that is simple to input voice and is less prone to repeated erroneous recognitions.

〔発明の概要〕 本発明の望ましい一実施態様においては、音声入力デー
タを予め登録されている音声パターンデータと比較照合
し最も近い音声パター(第1位の語)を認識結果として
出力するが、このとき、2番目の近い音声パターン(第
2位の語)も記憶しておき、誤認識され、操作者による
訂正コマンド入力後、再び音声入力した結果、第1位の
音声パターンが同じであれば、同じ誤認識が繰り返され
たと見なして、第2位の語を出力するようにする。
[Summary of the Invention] In a preferred embodiment of the present invention, voice input data is compared with pre-registered voice pattern data and the closest voice pattern (first word) is output as a recognition result. At this time, the second closest sound pattern (second place word) is also memorized, and if it is misrecognized and the operator enters a correction command and inputs the voice again, the first place sound pattern is the same. For example, it is assumed that the same misrecognition has been repeated, and the second-ranked word is output.

〔発明の実施例〕[Embodiments of the invention]

以下、本発明の一実施例を第1図により説明する。 An embodiment of the present invention will be described below with reference to FIG.

無線機移動局3と無線機固定局2との間は無線電波によ
って結合している。音声入出力装置1は無線機固定局2
の受信音声データを入力する音声認識装置6、音声認識
結果に判定し、その結果をガイダンスしたりする音声合
成装置7、および。
The radio mobile station 3 and the radio fixed station 2 are coupled by radio waves. Audio input/output device 1 is a radio fixed station 2
a speech recognition device 6 that inputs received speech data; a speech synthesis device 7 that judges the speech recognition result and provides guidance on the result;

これらを制御する制御部8、音声パターンデータ及び、
必要な情報を記憶する記憶部9より構成されている。音
声情報を入力する話者は、無線機移動局3を携帯してマ
スク4から音声でデータを入力する。音声入力結果は、
スピーカ(また畔イヤホン)5からのアンサーノ1?り
を聞<:T午により確認することができる。第2図は、
その他の実施例を示し無線機が無い点の、みが竿1図と
異なる。
A control unit 8 that controls these, audio pattern data, and
It is composed of a storage section 9 that stores necessary information. A speaker who inputs voice information carries a radio mobile station 3 and inputs data by voice through a mask 4. The voice input result is
Answer no 1 from speaker (also earphones) 5? This can be confirmed by listening to <:T. Figure 2 shows
This figure differs from Figure 1 only in that it shows another embodiment and does not include a radio.

話者が音声入力をし、それが誤認率された帰合の処理フ
ローを第3図に示訃、。
FIG. 3 shows a processing flow in which a speaker inputs voice and results in a misrecognition rate.

音声入力データが、[訂正」でないならば、通常常の処
理を行い、訂正フラグが1となって、いるかどうかを見
る。通常、r訂坦」が入力さ些なければ訂正フラグはO
となっていやので、そのまま認識結果の第1位の語を出
力し、同時に、第1位、第2位の語を記憶部へ第4.図
のように格納する。
If the audio input data is not [corrected], normal processing is performed to see if the correction flag is set to 1. Normally, the correction flag is set to O if "redan" is not input.
Therefore, the first word in the recognition result is output as is, and at the same time, the first and second words are transferred to the storage section in the fourth place. Store as shown.

話者が誤認識に気づき、前のみ力データを、訂正すべく
、「訂正」を入力す、るζ1.訂正フラグを1としてル
ーチンをぬ、ける。、次に、音声データ1を再入力する
と、すでに訂正、フラグが1、と4つてりるため、新し
い、データが、前;:、I!、認識さ、れたデータと―
じかどうかの判定に処、理がうつ、る。この時。
ζ1. The speaker notices the misrecognition and inputs "correction" to correct the previous force data. Set the correction flag to 1 and exit the routine. ,Next, when audio data 1 is re-inputted, there are already 4 corrections and 1 flags, so the new data is the previous;:,I! , recognized data and...
Processing and logic are involved in determining whether it is true or not. At this time.

新してデータが、前の誤認識、されたデニタと、同じで
あれば、そのデータをキャンセル鴫1.第2位の(5)
、 語を第1位の認識語として出力し、訂正フラグを0とし
てルーチンをぬける。また、新しいデータが前の誤認識
されたデータと異なっていれば、今度は正しく認識され
たものとして、第1位の語をそのまま認識結果として出
力し、訂正フラグを0としてルーチンをぬける。   
  ・    7、この実施例によれば、誤認識に対し
て新たな音声パターン、を増やすことなく、簡単確実に
正しい認識を可能とすることができる。
If the new data is the same as the previous incorrectly recognized data, cancel that data.1. 2nd place (5)
, outputs the word as the first recognized word, sets the correction flag to 0, and exits the routine. Furthermore, if the new data is different from the previous erroneously recognized data, it is assumed that it has been correctly recognized this time, and the first word is output as is as the recognition result, the correction flag is set to 0, and the routine is exited.
- 7. According to this embodiment, correct recognition can be easily and reliably achieved without increasing the number of new voice patterns for erroneous recognition.

なお、上記実施例においては、常に第2位の音声パター
ンを記憶マておき、誤認識時1.この記憶さ、れた(前
回第2位の)音声パターンに対応するアンサーバックを
行うものとしたが、今回第2位の音声パターンに対応す
るアンサーバックを行うこともでき、この場合には、操
作者が入力し直した音声であるから、前回第2位より誤
認識率を低重・舊に、前回と今回の第1位の音声パター
ンが二゛(できる。
In the above embodiment, the second highest voice pattern is always stored in memory, and in case of misrecognition, 1. The answerback corresponding to the memorized voice pattern (second place last time) was performed, but it is also possible to perform the answerback corresponding to the voice pattern ranked second this time. In this case, Since the voice was re-entered by the operator, it is possible to duplicate the voice pattern of the first place last time and this time, with a lower misrecognition rate than the second place last time.

致したことに・よって、第2のアンサーバックを行うも
のとしたが、第1および第、2位の音声パターン共に一
致したことを確認することもできる。
Therefore, a second answerback was performed, but it can also be confirmed that the first, second, and second place voice patterns all match.

〔発明の効果〕〔Effect of the invention〕

本発明によれば、話者が音声入力をする際、誤認識が発
生し、それを訂正し再入力をした時に、同じ誤認識が繰
り返されることが少ない音声入力装置を提供することが
できる。
According to the present invention, it is possible to provide a voice input device in which erroneous recognition occurs when a speaker inputs voice, and when the speaker corrects the erroneous recognition and re-inputs the voice, the same erroneous recognition is less likely to be repeated.

【図面の簡単な説明】[Brief explanation of drawings]

第1図は本発明による音声入出力装置の無線を用いた場
合の構成図、第2図は同じく有線を用いた場合の構成図
、第3図は本発明による音声入出力装置の要部を示すた
めの誤認識がおきた場合の入力データの訂正の処理フロ
ー、第4図は同じく記憶部のデータ格納の例を示す図で
ある。 1・・・音声入出力装置、4・・・マイク、5・・・ス
ピーカ、またはイヤホン、6・・・音声認識装置、7・
・・音声合成装置、8・・・制御部、9・・・記憶部。
Fig. 1 is a block diagram of the audio input/output device according to the present invention when using wireless, Fig. 2 is a block diagram when using wired as well, and Fig. 3 shows the main parts of the audio input/output device according to the present invention. FIG. 4 is a diagram showing an example of data storage in the storage unit. 1... Audio input/output device, 4... Microphone, 5... Speaker or earphone, 6... Voice recognition device, 7...
...Speech synthesis device, 8...Control unit, 9...Storage unit.

Claims (1)

【特許請求の範囲】[Claims] 1、音声入力を取り込み、該音声入力と、予め登録され
ている音声パターンデータとを比較照合し、認識結果を
得る音声入力部と、認識結果をアンサーバックする出力
部とを備えたものにおいて、入力された音声データに最
も近い第1の音声パターンに、対応する認識結果をアン
サーバックする手段と、アンサーバックされた出力に対
して操作者からの誤認識である旨の応答操作入力がなさ
れたとき次回の音声入力により音声認識結果と前回の上
記記憶内容を照合する手段と、この照合結果が一致した
とき次に近い第2の音声パターンデータに対応するアン
サーバック処理を実行する手段とを設けた音声入力装置
1. A device comprising a voice input unit that takes in voice input, compares and matches the voice input with pre-registered voice pattern data, and obtains a recognition result, and an output unit that answers back the recognition result, A means for answering back a recognition result corresponding to a first voice pattern closest to the input voice data, and a response operation input from an operator to the effect of erroneous recognition in response to the answered back output. means for comparing the voice recognition result with the previous memory content according to the next voice input; and means for executing answerback processing corresponding to the next closest second voice pattern data when the comparison results match. voice input device.
JP60072637A 1985-04-08 1985-04-08 Voice input device Pending JPS61231629A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
JP60072637A JPS61231629A (en) 1985-04-08 1985-04-08 Voice input device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
JP60072637A JPS61231629A (en) 1985-04-08 1985-04-08 Voice input device

Publications (1)

Publication Number Publication Date
JPS61231629A true JPS61231629A (en) 1986-10-15

Family

ID=13495100

Family Applications (1)

Application Number Title Priority Date Filing Date
JP60072637A Pending JPS61231629A (en) 1985-04-08 1985-04-08 Voice input device

Country Status (1)

Country Link
JP (1) JPS61231629A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH11109989A (en) * 1997-10-02 1999-04-23 Toyota Motor Corp Speech recognition device
JPH11143487A (en) * 1997-11-11 1999-05-28 Osaka Gas Co Ltd Method and device for converting voice to character
JP2017029306A (en) * 2015-07-30 2017-02-09 井関農機株式会社 Automatic rice washing and cooking machine

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH11109989A (en) * 1997-10-02 1999-04-23 Toyota Motor Corp Speech recognition device
JPH11143487A (en) * 1997-11-11 1999-05-28 Osaka Gas Co Ltd Method and device for converting voice to character
JP2017029306A (en) * 2015-07-30 2017-02-09 井関農機株式会社 Automatic rice washing and cooking machine

Similar Documents

Publication Publication Date Title
WO2021159688A1 (en) Voiceprint recognition method and apparatus, and storage medium and electronic apparatus
JP3168033B2 (en) Voice telephone dialing
EP0655732A2 (en) Soft decision speech recognition
JPS61231629A (en) Voice input device
CA1239478A (en) Method and apparatus for use in interactive dialogue
JPS597998A (en) Continuous voice recognition equipment
JPH0432900A (en) Sound recognizing device
WO1994002936A1 (en) Voice recognition apparatus and method
JPS59144946A (en) Controlling system of voice typewriter
JPH11109989A (en) Speech recognition device
JPS59195299A (en) Sepecific speaker&#39;s voice recognition equipment
JP2001175279A (en) Speech recognizing method
JPH10301595A (en) Voice recognition and response device
JPS5962900A (en) Voice recognition system
JPS5915990A (en) Voice recognition system
JPH01197795A (en) Voice recognizing device
JPH02149900A (en) Voice recognizing and answering device
JPS5946695A (en) Voice recognition system
JPS5917597A (en) Voice recognition system
JPH02275497A (en) Voice recognition device
JPS5917598A (en) Voice recognition system
JPS59147397A (en) Voice recognition control system
JPS5945499A (en) Voice recognition system
JPS61235900A (en) Voice recognition method
JPH01191200A (en) Voice recognizing device