JP2000356999A - Device and method for inputting command by voice - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve the flexibility for a voice request. SOLUTION: The device is provided with a voice signal converting section 2 which conducts voice recognition by comparing and collating a voice request that is pronounced by an operator for an operation object body and plural keywords that are used to specify commands to be provided to the body and outputs one or plural keywords having high degree of suitability for the request as voice recognition candidates, an interactive example database 4 which stores a keyword string that are made by combining plural keywords in a prescribed order for every command as an interactive example, and a command interpreting section 3 which outputs a command corresponding to the keyword string to the operation object body when the keyword string that matches with the keyword inputted from the section 2 is detected from the database 4.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an apparatus and a method for inputting a command by voice, and more particularly to a technique for recognizing a voice request of an operator as a command to a plant using a voice recognition technique.

[0002]

2. Description of the Related Art JP-A-5-137181 describes a plant control system for operating a plant by voice input. This plant control system realizes automatic operation of a plant by recognizing a voice input as an operation command for a plant using a voice recognition technology, reading an operation method corresponding to the operation command from a database.

On the other hand, Japanese Patent Application Laid-Open No. 5-290101 discloses a continuous speech recognition processing technique in which the recognition rate of an uttered sentence is improved. This technology evaluates plant data using a plant function hierarchical model, operation knowledge base, and dialog history knowledge base to narrow down the attention area of the operator and determine the purpose and content of the operator's utterance. At the same time, the purpose and content of the utterance of each of the plurality of recognition sentence candidates generated by the speech recognition unit are extracted, and the purpose and content of the utterance determined by narrowing down the range of interest are compared. Make a decision.

[0004]

However, in such prior art, a driver is forced to request a standard pattern of voice in order to secure a desired voice recognition rate. In other words, because of the lack of flexibility for voice requests,
Practical usability was poor. Further, the prior art is insufficient in that the voice request of the operator is unilaterally received and recognized, and thus the voice request is more reliably recognized. In particular, when a command input device based on such a voice recognition technique is applied to a plant, it is necessary to accurately recognize a voice request of an operator and generate a command that more accurately reflects the voice request of the operator.

The present invention has been made in view of the above-mentioned problems, and has the following objects. (1) Improve flexibility for voice requests. (2) Improve practical usability. (3) Recognize the voice request more reliably.

[0006]

In order to achieve the above-mentioned object, according to the present invention, as a first means relating to a command input device by voice, a voice request for an operation object pronounced by an operator and an operation object are provided. Voice recognition means for performing voice recognition by comparing and collating with a plurality of keywords for specifying a command to be provided, and outputting one or a plurality of keywords having high relevance to the voice request as voice recognition candidates; A dialogue case storage unit for storing a keyword sequence in which keywords are combined in a predetermined order as a dialogue case for each command, and a keyword sequence matching the keyword input from the voice recognition unit is detected from the dialogue case storage unit. And a command interpreter for outputting a command corresponding to the operation target to the operation target.

The second related to the command input device by voice
In the above-mentioned first means, the command interpreting unit according to the first means, when there is a shortage of keywords as speech recognition candidates in the keyword sequence as a dialogue example, replaces the missing keyword not input from the speech recognition means. A means is employed in which complement is performed based on the keyword input immediately before, and a command corresponding to the keyword string obtained by the complement is output to the operation target.

The third related to the command input device by voice
In the first means, in the first means, when a keyword sequence in which the order of keywords as speech recognition candidates is replaced with a keyword sequence as a dialogue example is input from the speech recognition means, And outputting a command corresponding to the keyword string obtained by transposing the order to the operation target.

The fourth related to the command input device by voice
Means, in any one of the first to third means, further comprising a voice synthesizing means for voice-synthesizing a voice message to the operator based on an instruction of the command interpreting unit;
If the keyword as a speech recognition candidate does not match the keyword string as the dialogue example,
A means for instructing the voice synthesizing means to synthesize a voice message urging the operator to input a voice necessary for specifying the command is employed.

Fifth related to command input device by voice
In the fourth means, the voice recognition means is configured to limit a range of a keyword to be used for voice recognition of a voice request in accordance with an instruction of the command interpretation unit,
The command interpreting unit outputs, to the voice recognizing means, an instruction to limit the range of the keyword to those related to an operator's response to the voice message when the voice synthesizing means causes the voice message to be synthesized. adopt.

On the other hand, according to the present invention, as a first means relating to a command input method by voice, a keyword sequence in which a plurality of keywords for specifying a command to be provided to an operation target are combined in a predetermined order is interacted for each command. A step of storing in advance as a case, and performing voice recognition by comparing a voice request pronounced by an operator with the keyword, and outputting one or a plurality of keywords having high relevance to the voice request as voice recognition candidates. And a step of comparing and collating the keyword as a voice recognition candidate with a keyword string as a dialogue example, and outputting a command corresponding to the keyword string to the operation target when both match. .

The second related to the command input method by voice
In the first means, when the keyword as the speech recognition candidate is insufficient for the keyword sequence as the dialogue example, the keyword is complemented based on the keyword input immediately before. Means of outputting a command corresponding to the obtained keyword string to the operation target is adopted.

The third related to the command input method by voice
In the first means, when a keyword in which the order of the keywords as the speech recognition candidates is exchanged with respect to the keyword sequence as the dialogue example is input, the keyword obtained by transposing the order of the keywords is obtained. Means of outputting the command corresponding to the column to the operation target is adopted.

The fourth related to the command input method by voice
In any one of the first to third means, if the keyword as a speech recognition candidate does not match the keyword string as a dialogue example, the operator inputs a voice necessary for specifying a command by a voice message. Adopt means to prompt.

Fifth related to a command input method by voice
Means for limiting the range of keywords used for voice recognition of a voice request to those related to the operator's response to the voice message in the fourth means. adopt.

[0016]

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram showing an embodiment of a voice command input device and method according to the present invention; The present embodiment is an application of the present invention to plant operation, and more specifically relates to a case where a command is input to a plant control device that is an operation target based on a voice request of an operator. .

FIG. 1 is a block diagram of a voice command input device (hereinafter, voice command input device) according to the present embodiment. As shown in the figure, the voice command input device A includes a microphone 1, a voice signal conversion unit 2 (voice recognition unit), a command interpretation unit 3, a dialogue case database 4 (dialog case storage unit), and a Japanese synthesis output unit. 5 (speech synthesis means) and a speaker 6.

The microphone 1 converts a voice request uttered by the operator into an operator voice signal. In the case of the present embodiment, the operator is a plant operator who monitors the operation of the plant via the plant control device, and the voice request uttered by the plant operator is, for example, the operation state of various devices constituting the plant. A device that instructs the control device to display, a device that instructs a change in the operation state of the various devices, and the like may be considered.

The voice signal conversion unit 2 is a kind of voice recognition engine. The voice signal conversion unit 2 recognizes an operator voice signal (voice request) supplied from the microphone 1, and sends a keyword sequence as a recognition candidate to the command interpretation unit 3. Output. The voice signal conversion unit 2 determines the degree of matching between the operator voice signal and a keyword stored in advance, and outputs a keyword having a high degree of matching together with the value of the degree of matching to the command interpreting unit 3 as a recognition candidate. Further, the audio signal conversion unit 2 records a plurality of keywords in advance in a plurality of groups and records each keyword in the group specified by the dialogue status selection signal supplied from the command interpretation unit 3 as described later. It is configured to compare and refer to the keyword with the operator voice signal.

Here, the keywords stored in the audio signal conversion unit 2 will be specifically complemented. This keyword is composed of a keyword for specifying the “operating subject” and a keyword for specifying “operation content” for the operating subject. For example, when the operation state of a certain device is displayed on the plant control device, the operator issues a voice request specifying at least “device” as the operation subject and “display” as the operation content.

The keyword for specifying the "operator" and the keyword for specifying the "operation content" are registered in advance in the audio signal converter 2, and the keyword and the audio request are registered. Voice recognition of the voice request of the operator is performed based on the comparison and collation. When an operator voice signal is input based on a voice request, the voice signal conversion unit 2 extracts a keyword for specifying the “operating subject” and a keyword for specifying “operation content”,
A keyword string composed of these two keywords is output to the command interpreter 3. Such a keyword string is used as a recognition speech recognition candidate together with the value of the matching degree by the command interpreter 3.
Or one or more. Note that such keywords are appropriately set according to the content of a command that can be input to the plant control device, that is, according to the operation target.

Further, the above-mentioned keywords are registered by being grouped into several types according to the content of the voice request of the operator (that is, the dialogue situation). For example, when the operator makes a voice request to the plant control device to display the operating state of the equipment, the content of the voice request is clearly different between when the plant is operating normally and when it is abnormal. That is, in a dialogue situation where a voice request related to normal operation of the plant is input, a voice request related to abnormal plant is not input from the operator.

As will be described later, in a dialogue situation where the voice command input device A inquires the operator about the execution confirmation of the voice request or the request confirmation in response to the voice request from the operator, a general voice request dialogue is performed. A voice limited to the situation will be input from the operator. In a case where the voice command input device A accepts a general voice request from an operator (interaction situation),
Since it is unknown what voice request is input, it is necessary to use all the keywords registered in the voice signal converter 2 for voice recognition of the voice request.

However, in a dialogue situation in which it is known that only a limited content of voice is input, a keyword to be used for voice recognition is narrowed down according to the above-described limited voice, so that the recognition rate is reduced. It is possible to realize improvement and reduce the load on the audio signal conversion unit 2 (speeding up processing, etc.).

In the present embodiment, in consideration of such a difference in the content of the voice request depending on the dialogue situation, the keywords are pre-divided into several groups according to the dialogue situation and registered in the voice signal converter 2. The voice signal converter 2 reduces the load on the voice signal converter 2 by comparing and referencing a keyword of a specific group specified by the dialogue status selection signal input from the command interpreter 3 with the operator voice signal. The aim is to improve the recognition rate of voice requests.

The command interpreting unit 3 interprets each keyword string sequentially input from the voice signal converting unit 2 as a command to be provided to the plant control device by comparing and collating with a dialogue example stored in the dialogue case database 4 in advance. ,
This command is output to the plant control device.
If the command interpretation unit 3 cannot interpret each keyword string as a command, it instructs the Japanese synthesis output unit 5 to inquire about a keyword (missing keyword) necessary for interpreting the command string. In addition,
The detailed operation of the command interpreter 3 will be described later.

The dialogue case database 4 stores a plurality of keyword strings corresponding to commands that can be provided to the plant control apparatus as dialogue cases. As described above, the plant control device can display the operating state for each device constituting the plant, and a command for displaying the operating state for each device is prepared as a command. In response to such a command, the dialogue case database 4 includes a keyword sequence in which keywords for specifying the “device” as the operation subject and keywords for specifying the operation content of “display” are combined in a normal order. Is stored as a dialogue example.

Here, when an operation for a certain operation subject is provided to the plant control apparatus as a command, it is normal that the operation content follows the operation subject as the voice request. For example, when it is desired to display the operating state of the pump as the device on the plant control device, the operator issues a voice request to "display (operate) the pump (operation subject)" as normal Japanese. In this case, various variations can be considered for the particles and suffixes. In normal Japanese, however, the voice request is such that at least “pump”, which is the operation subject, and “display” indicating the operation content follow. The dialogue case database 4 stores keywords having a normal (typical) array as a Japanese voice request as a dialogue case for each command.

As described above, a plurality of keyword strings corresponding to commands that can be provided to the plant control device are stored in the dialog case database 4 as dialog examples, and these keyword strings, that is, each command, require execution confirmation. It is stored together with additional information indicating whether the command is a command. According to the supplementary information, for example, a command for instructing the operation state of various devices to be displayed on the plant control device is set to a command that does not require execution confirmation, and a command for instructing a change in the operation state of various devices is set to a wrong plant Command is not allowed, so it is set to a command that requires execution confirmation.

The Japanese synthesis output unit 5 generates a re-listening signal for asking the operator what the missing keyword is based on an inquiry instruction of the “missing keyword” input from the command interpreting unit 3 and the like. ,
The re-listening signal is output to the speaker 6. The speaker 6 converts the re-listening signal into voice and notifies the operator of voice.

Next, the operation of the voice command input device A thus configured will be described in detail with reference to the state transition diagram shown in FIG. In this state transition diagram, processes S01 to S06 framed by an ellipse indicate processes by the operator, and processes S11 to S15 framed in a square indicate processes by the voice command input device A.

First, the voice command input device A changes the operation mode from the sleep mode to the active mode when, for example, a specific start instruction voice is recognized by the voice signal converter 2, and thereafter, the specific end instruction is given. Until the voice is recognized, the voice input from the operator to the microphone 1 is recognized and interpreted as a voice request to the plant control device.

In the initial state immediately after the transition to the active mode, when a voice request for prompting the display of the operating state of a specific device is input to the microphone 1 from the operator, the voice request is transmitted from the microphone 1 as an operator voice signal. The speech signal is input to the speech signal conversion unit 2 and is converted into a keyword string which is a speech recognition candidate. In this case, the voice signal conversion unit 2 compares and verifies the voice request with a keyword stored in advance, and outputs the keyword request to the command interpretation unit 3 in order from a keyword string having a higher degree of matching.

For example, when the operator inputs a voice request to "display a pump" to the microphone 1, the voice signal conversion unit 2 outputs "pump" and "display" as keywords in advance. A keyword string consisting of “+ display” is output to the command interpreter 3. Here, if the operator inputs a voice request of “display, pump” to the microphone 1, the arrangement of individual keywords is “display + pump” transposed with respect to the previous “display pump”. Is output to the command interpreter 3.

The command interpreter 3 searches the dialogue case database 4 for the keyword sequence input from the voice signal converter 2 in this manner, and extracts a dialogue case that matches the keyword sequence. When a keyword string is input as the request S01, the command interpreting unit 3 firstly determines whether the keyword string is a request that can be sufficiently interpreted and does not require confirmation, or is sufficiently interpretable but important. It is determined whether the request is a request that needs to be confirmed before execution because it is an operation, or whether it is an unclear or ambiguous request that cannot be interpreted.

Then, the command interpreter 3 sends the request S
When a request for which the request 01 is not required to be confirmed, that is, for the request S01, a voice example that matches the keyword string input from the voice signal conversion unit 2 is detected by searching the dialogue case database 4, a transition is made along the arrow b1. The request S01 is executed S11. That is, the command interpreter 3
Generates a command corresponding to the request S01 and outputs the command to the plant control device.

When the execution S11 for one request S01 is completed in this way, the command interpreting unit 3 accepts a new request S01 from the operator by making a transition along the arrow b2, or along the arrow b3. By making the transition, the process shifts to a standby state for receiving a correction request S05 for the request S01 that has just been executed S11.

Here, for example, in the case where a dialogue example consisting of “pump + display” is registered in the dialogue case database 4, the operator may call “display, pump” and “pump” and “display”. When a voice request in which is transposed is input, the command interpreter 3 outputs “display + pump”
Is determined to be equivalent to the above “pump + display” and the request S
Execute 01 in S11. That is, assuming a voice request in which such a transposition is performed, the dialogue case database 4 stores “display +
A dialogue example of “pump” is stored in advance, and the command interpreting unit 3 executes the request S01 based on the dialogue example.

On the other hand, the command interpreter 3 sends the request S01
If it is determined that confirmation is necessary before execution by searching for a request requiring confirmation, that is, by searching the dialogue case database 4, the request S01 is made by transiting along the arrow b5.
Is performed on the execution confirmation S12. For example, if the operator issues a voice request regarding a change in the operating state of the pump (equipment) such as “set the pump output to 25%”, the command interpreting unit 3 will never operate the plant incorrectly. Since it is necessary to avoid this, after asking "Set the pump output to 25%, are you sure?", The execution confirmation S12 is performed. In this case, the command interpreter 3
Instructs the Japanese synthesis output unit 5 to perform the above-mentioned re-listening and notifies the speaker 6 from the speaker 6.

When only one of "pump" and "display" is extracted by the audio signal converter 2, the command interpreter 3 comprises either "display" or "pump". The keyword will be entered. In such a case, the command interpreting unit 3 attempts to interpret the voice request of the operator by complementing the keyword that was not input from the voice signal converting unit 2 (the missing keyword). If the request obtained as a result does not require execution confirmation, execution S11 is performed by transitioning along arrow b1. If the request requires execution confirmation, transition is performed along arrow b5. An execution confirmation S12 is performed for the request S01.

For example, following a voice request of "display a pump", if only a keyword having a device name of "tank" different from the pump is input to the command interpreting unit 3, the command interpreting unit 3 will The missing keyword “display” is identified based on a comparison with the keyword column “pump + display”.

If the above complementation cannot be performed sufficiently, the command interpreting section 3 performs request confirmation S13 by making a transition along the arrow b15. For example, when only the keyword “setting” is input to the command interpreting unit 3, the command interpreting unit 3 cannot compensate for all the missing keywords based on the keyword string “pump + display” input immediately before. A request confirmation S13 for asking "How many pumps should be set?" At this time, the command interpreter 3 asks "How many pumps should be set?"
The voice synthesizing output unit 5 is instructed to generate a re-listening voice signal corresponding to the voice message, and the voice message is notified from the speaker 6.

Further, in response to such a request confirmation S13, when the operator inputs a selection / correction S02 of "25%" (in the case of the transition by the arrow b16), the command interpreting section 3 sets "Pump + 25%". The request S01 is interpreted based on the keyword string “+ setting”. In this case, since the request S01 is a request that requires execution confirmation, the arrow b
The execution confirmation S12 is performed by making a transition along the line 19.

In response to the execution confirmation S12, the operator can take an action as to agree with the interpretation of the voice command input device A (permission S03) or disagree with the interpretation of the voice command input device A ( Either a non-permission S04), a correction request S05, or a cancellation of the request S06 is made. In such a dialogue situation, the content of the voice input by the operator to the voice command input device A is limited as compared with the case where the operator's request S01 is initially received as described above.

That is, in an interactive situation in which the execution confirmation S12 is performed for the request S01 once input, the voice input to the voice command input device A by the operator is limited to any of the above four. Therefore, in such a dialogue situation, the command interpreting section 3 outputs a dialogue situation selection signal for specifying a group associated with these four correspondences to the audio signal conversion section 2. When an operator's voice signal corresponding to the execution confirmation S12 is input from the operator, the voice signal conversion unit 2 compares the keyword of the group specified by the dialogue status selection signal with the operator's voice signal, thereby obtaining a keyword having a high degree of matching. Is extracted and output to the command interpreter 3.

For example, when a voice for permitting interpretation of the voice command input device A S03 is input from the operator as a response to the execution confirmation S12 (arrow b6), the command interpreting section 3 makes a transition along the arrow b7. As a result, a command based on its own interpretation is generated and output to the plant controller, that is, execution S11.

When the correction request S05 corresponding to the request S01 is input (arrow b8), the command interpreter 3 inputs the keyword input from the voice signal converter 2 for the correction request S05 to the dialogue case database 4. If it exists as a case, that is, if it can be executed, a transition is made along the arrow b4 to generate and execute S11 a command corresponding to the correction request S05. In this case, the execution confirmation S12 may be required again for the correction request S05 (arrow b9).

When the voice of the disapproval S04 corresponding to the request S01 is input (arrow b10), the command interpreter 3
By making a transition along the arrow b11, a correction promotion S14 is requested of the operator. In the correction promotion S14, the command interpreting unit 3 instructs the Japanese generation output unit 5 to generate a re-listening audio signal for inquiring how to correct. Then, when the regenerated audio signal is generated, the Japanese generation output unit 5 outputs the signal to the speaker 6 to notify the operator.

Here, in the dialogue situation where the operator requests the correction promotion S14 in this way, the operator's response is either the correction request S05 to the request S01 or the suspension S05 of the request S01. Therefore, in such a dialogue situation, the command interpreter 3 outputs a dialogue situation selection signal designating a group related to the correction request S05 and the cancellation S05 to the voice signal converter 2 to be used for recognition of the voice of the operator. The keywords are limited to those suitable for the dialogue situation.

When a correction request S05 is input as a response to the correction promotion S14 (transition by arrow b13),
In the same manner as in the case of the correction request S05 for the execution confirmation S12 described above, if the execution confirmation is unnecessary, the process transits along the arrow b4 to execute the correction request S05. By making a transition along the arrow b9, the execution confirmation S12 is again requested to the operator.

On the other hand, the command interpreter 3 sends the request S01
If it is determined that is meaningless or ambiguous, the arrow b
The request confirmation S13 is performed by transitioning along the line 15. In such a case, for example, if no dialogue example that matches the keyword input from the voice signal conversion unit 2 for the request S01 is detected by searching the dialogue case database 4, the keyword input from the voice signal conversion unit 2 is complemented. Failed to detect a suitable dialogue example, or if a suitable dialogue example was detected but its relevance was low, or multiple dialogue cases were detected and there was no significant difference in their relevance. And so on.

In such a situation, as an action of the operator, selection / correction S02 of request S01 or suspension S06 of request S01 can be considered. Here, when the selection / correction S02 corresponding to the request confirmation S13 is input from the voice signal conversion unit 2 (arrow b16), when the selection / correction S02 is executable, that is, the selection / correction S02 is performed. If a dialogue case that matches the keyword input from is detected by searching the dialogue case database 4, the execution confirmation S12 is performed by transiting along the arrow b19 when execution confirmation is necessary, and when the execution confirmation is unnecessary. The request S01 is directly executed S11 by transiting along the arrow b17.

On the other hand, if the selection / correction S02 is not executable, that is, even if the selection / correction S02 is performed, the request S01 is not made.
Is unknown or ambiguous (arrow b18), the command interpreter 3 performs the request confirmation S13 again.

Further, in response to the request confirmation S13, the request S01
When the stop S06 is instructed by the operator, that is, when the transition is made along the arrow b20 (also when the transition is made along the arrows b12 and b14), the command interpreting section 3 makes the transition along the arrow b21. As a result, the dialog relating to the previous request S01 is stopped (dialog stop S15). Then, when the dialog stop S15 is determined, the command interpreter 3 transitions along the arrow b21 to enter a standby state waiting for a new request S01 to be input.

As described above, in the command input device of the present embodiment, the keyword sequence related to the voice request obtained as a result of the voice recognition of the voice signal converter 2 is stored in the dialog case database 4 in the dialog case (keyword sequence). A command corresponding to the voice request is generated and output to the plant control device (that is, executed) only when the request does not need to confirm the intention of the operator for execution.
If the request requires confirmation of the operator's intention, the request is executed after execution confirmation S12. Further, when it is determined that the request is unclear or ambiguous, a correct request is derived by request confirmation and then executed or executed after execution is confirmed.

Therefore, according to the present embodiment, since the sound content of the voice request is flexibly handled, the practical usability is improved, and a command reflecting the voice request of the operator more accurately is transmitted to the plant control. Since the data can be output to the device, the reliability of the command input device can be improved.

The command input device according to the present embodiment
In the case of execution confirmation S12 and correction promotion S14, etc., by outputting a dialogue state selection signal to the voice signal conversion unit 2 according to the dialogue state at that time, the contents of the keyword used for recognition of the voice request of the operator are answered by the operator. Is limited to those related to (dynamic). Therefore, it is possible to reduce the load on the voice signal conversion unit 2 and to improve the voice recognition rate for the voice request, so that the reliability can be further improved.

Although the present invention is applied to the case where the voice request of the operator is input as a command to the plant control device which is the operation target in the above embodiment, the present invention is not limited to this. It is not something to be done. INDUSTRIAL APPLICABILITY The present invention is applicable to a case where a command based on a voice request is provided to various operation objects other than the plant control device.

Further, in the above embodiment, the keyword string including the keyword for specifying the operation subject and the two keywords for specifying the operation content is used, but the present invention is not limited to this. Depending on the operation target, a command may be specified by a keyword string including three or more keywords. Further, the state transition diagram shown in FIG. 2 is merely an example, and it is conceivable to set another state transition when the present invention is applied to plant operation.

[0060]

As described above, the voice command input device and method according to the present invention have the following effects.

(1) According to the first or sixth aspect of the invention, a keyword sequence in which a plurality of keywords for specifying a command to be provided to an operation target are combined in a predetermined order is used as a dialog example for each command. A voice request that is stored in advance and is spoken by an operator is subjected to voice recognition by comparing and collating with the above-mentioned keyword, and one or more keywords having high relevance to the voice request are set as voice recognition candidates, and are set as the voice recognition candidates. Is compared with the keyword string as a dialogue example, and if the two match, the command corresponding to the keyword string is output to the operation target, so the command that more accurately reflects the voice request of the operator is plant controlled. It can be output to the device. Therefore, the reliability of the voice command input device can be improved.

(2) According to the second or seventh aspect of the present invention, when there is a shortage of keywords as speech recognition candidates in a keyword sequence as a dialogue example,
Completion is performed based on the keyword input immediately before, and a command corresponding to the keyword string obtained by the completion is output to the operation target, so that it is possible to flexibly respond to the pronunciation content of the voice request. Therefore, practical usability of the command input device can be improved. Further, in this case, the operator can omit a part of the keyword when the voice request is continuously input to the command input device by voice, so that the load on the operator can be reduced. Is possible.

(3) According to the invention as set forth in claim 3 or claim 8, when a keyword in which the order of keywords as voice recognition candidates has been exchanged with respect to a keyword sequence as a dialogue example is input, the keyword Since the command corresponding to the keyword sequence obtained by transposing the order is output to the operation target, it is possible to flexibly respond to the pronunciation content of the voice request, and therefore, the practical use of the command input device by voice becomes practical. Can be improved.

(4) According to the invention as set forth in claim 4 or claim 9, when a keyword as a speech recognition candidate does not match a keyword sequence as a dialogue example, a speech necessary for specifying a command is input. Since the voice message prompts the operator, it is possible to flexibly respond to the pronunciation content of the voice request, so that the practical usability of the voice command input device can be further improved.

(5) According to the invention described in claim 5 or claim 10, the range of keywords used for voice recognition of a voice request is limited to those related to the response of the operator to the voice message. Can be reduced, and the voice recognition rate for voice requests can be improved, so that reliability can be further improved.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a functional configuration of a voice command input device according to an embodiment of the present invention.

FIG. 2 is a state transition diagram showing an operation of the voice command input device according to one embodiment of the present invention.

[Explanation of symbols]

 A voice command input device 1 microphone 2 voice signal converter (voice recognition means) 3 command interpreter 4 dialogue case database (dialogue case storage means) 5 Japanese synthesis output unit ( Voice synthesis means) 6. Speaker

 ────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Yuji Matsuura 3-1-1, Toyosu, Koto-ku, Tokyo AA01 BB09 BB11 CC07 CX06 JJ05 JJ15 9A001 HH17 HH18 JJ48 KK54 LL09

Claims (10)

[Claims] 1. A voice request for an operation target pronounced by an operator is compared with a plurality of keywords for specifying a command to be provided to the operation target. Voice recognition means (2) for outputting one or a plurality of keywords having a high keyword as voice recognition candidates, and dialogue case storage means (4) for storing a keyword sequence in which a plurality of the keywords are combined in a predetermined order as a dialogue case for each command. A command interpreting unit that outputs a command corresponding to the keyword string to the operation target when a keyword string matching the keyword input from the voice recognition unit (2) is detected from the dialogue case storage unit (4). 3) A command input device by voice, comprising: 2. A command interpreting unit (3) which, when there is a shortage of keywords as speech recognition candidates in a keyword sequence as a dialogue example, deletes a missing keyword not input from the speech recognition means (2). The command input device using voice according to claim 1, wherein the command is complemented based on a keyword input immediately before, and a command corresponding to the keyword string obtained by the complementation is output to the operation target. 3. A command interpreting unit (3) which, when a keyword sequence in which the order of keywords as speech recognition candidates is replaced with a keyword sequence as a dialogue example is input from the speech recognition unit (2), The command input device by voice according to claim 1, wherein a command corresponding to the keyword sequence obtained by transposing the order is output to the operation target. 4. A voice synthesizing means (5) for voice-synthesizing a voice message to an operator based on an instruction of a command interpreting unit (3), wherein the command interpreting unit (3) performs voice recognition on a keyword sequence as a dialogue example. 4. The apparatus according to claim 1, wherein when the keywords do not match, the voice synthesizing unit is instructed to synthesize a voice message urging the operator to input a voice necessary for specifying the command. A command input device by voice described in 1. 5. The voice recognition means (2) is configured to limit a range of a keyword to be used for voice recognition of a voice request in accordance with an instruction of the command interpretation unit (3), and the command interpretation unit (3) performs voice synthesis. When the means (5) synthesizes a voice message by voice, outputting an instruction to the voice recognizing means (2) to limit the range of the keyword to those related to the response of the operator to the voice message. The voice command input device according to claim 4. 6. A step of storing, in advance, a keyword sequence in which a plurality of keywords for specifying a command to be provided to an operation target in a predetermined order as a dialogue example for each command; A step of performing speech recognition by comparing and collating with a keyword and outputting one or more keywords having a high degree of relevance to the speech request as speech recognition candidates; a keyword sequence as the speech recognition candidate and a dialogue example And outputting a command corresponding to the keyword string to the operation target when the two match, and a command input method by voice. 7. When a keyword string as a speech recognition candidate is insufficient for a keyword string as a dialogue example, complement is performed based on a keyword input immediately before, and a keyword string obtained by the complementation is added to the keyword string obtained by the complementation. 7. The command input method by voice according to claim 6, wherein a corresponding command is output to the operation target. 8. When a keyword in which the order of keywords as voice recognition candidates is exchanged with respect to a keyword string as a dialogue example is input, a command corresponding to a keyword string obtained by transposing the order of keywords is output. 7. The method according to claim 6, wherein the command is output to an operation target. 9. The method according to claim 6, wherein when a keyword as a voice recognition candidate does not match a keyword string as a dialogue example, an operator is prompted by a voice message to input a voice necessary for specifying a command. 8. The command input method by voice according to any one of 8. 10. The method according to claim 9, wherein when a voice message is provided to the operator, the range of keywords used for voice recognition of the voice request is limited to those related to the response of the operator to the voice message. Command input method by voice.