JPH07219591A - Voice processing device and method thereof - Google Patents

Abstract

PURPOSE:To make the resumption of the result of a malfunction easy by estimating and issuing an action invalidating the executed result of a voice command executed previously when the designation of the invalidation of the executed voice command is inputted. CONSTITUTION:When it is decided that a voice is inputted, a voice recognizing procession is performed by a voice recognition part 102. Further, whether the recognized voice command is of a completion instruction or not is decide and when it is decided to be 'no', whether the recognized voice command is of an invalidation instruction or not is decided. At this time, in the case it is decided to be an invalidation command (in the case of 'yes'), the estimation of an invalidating action for invalidating a command executed just before recorded in a command recording part 107 is performed in an invalidating action estimating part 105 to send the action to an action issuing part 108 and a command corresponding to the action is registered in a command recording part 106. Then, the invalidating action is issued from the action issuing part 108 to execute the invalidation.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice processing apparatus and method as a front end of an information processing apparatus, which replaces an input voice with an action such as a corresponding command sequence and issues the voice.

[0002]

2. Description of the Related Art In a conventional system having a voice input function, a dictionary at the time of recognition is dynamically switched to a vocabulary corresponding to a function which can be executed at that time, and misrecognition is performed while increasing the number of vocabularies which can be recognized as a whole. There are those that try to reduce the operability and improve the operability.

In order to prevent erroneous operation due to erroneous recognition of voice, measures are taken such as notifying the user of the result of voice recognition of a command input by voice and executing the command after reconfirmation by the user. There is also.

[0004]

However, when a voice command is erroneously recognized and executed, if the vocabulary that can be recognized is dynamically switched, a situation in which the voice command originally intended by the user cannot be recognized and executed occurs. ,
Even if the user repeatedly inputs a voice command, it may be invalid. At this time, the user must take action to restore the condition before executing the misrecognized command, and the action depends on the executed command, and the result of the system operation due to the misrecognition must be restored. Is a heavy burden on the user.

For example, if the user selects item B on the menu screen by mistake while trying to select item A, the screen is already in the submenu B, and There is no item A in. To reselect item A, you must exit the B submenu screen and return to the previous menu. Also, if you try to miniaturize a window while an application is running and you close it, you will not be able to miniaturize the window without opening the file again. In addition, when a graphic is moved while the screen is being edited with drawing software and the graphic is deleted, in order to move the graphic, the deleted graphic must be restored.

However, in order to prevent such malfunctions,
There is a method in which the result of the voice recognition is specified to the user and executed again after the user reconfirms, but this does not require the user to perform the two operations of input and confirmation, which is troublesome.

Further, even if the erroneous recognition is judged, if the user rephrases, the erroneously recognized vocabulary is included in the recognition target vocabulary, so that the erroneous recognition may be repeated.

[0008]

SUMMARY OF THE INVENTION In order to solve the above-mentioned conventional problems, the present invention records the execution of a voice command executed immediately before, and specifies the invalidation of the executed voice command. When input, it estimates the action that invalidates the execution result of the previously executed voice command, and issues this action to facilitate the repair of the malfunction result. It is intended to reduce the burden on the user by reducing the adverse effect of malfunction while avoiding the trouble of twice.

Further, with respect to the voice command invalidated as the erroneous recognition, the repetition of the erroneous recognition is prevented by removing it from the recognition target vocabulary of the immediately following voice command.

[0010]

Embodiments of the present invention will now be described in detail with reference to the drawings.

FIG. 1 is a diagram showing a block configuration of a voice processing apparatus according to an embodiment of the present invention.

In FIG. 1, 101 is a voice input unit, 1
Reference numeral 02 is a voice recognition unit that recognizes the voice input from 101, 103 is a recognition target vocabulary selection unit that switches the current recognition target vocabulary, and 104 is a reading of the recognized word, a voice feature, and a voice command of the word. Is a word dictionary that records and holds actions to be executed as, an invalidation action estimation unit 105 that determines an action to be performed when an invalid instruction is given, and a recognized voice command is a normal voice command. Or a voice command such as “different”, “cancel” or “invalid” that invalidates the immediately preceding voice command, and passes an action corresponding to it to the action issuing unit (108), 1
Reference numeral 07 is a command recording unit that records the command sent from the command determination unit 106, and 108 is an action issuing unit that issues the action of the voice command.

Next, the operation of the speech processing apparatus configured as described above will be described with reference to the flowchart of FIG.

First, in S201, it is determined whether or not a voice input is made from the voice input unit 101. If not, loop until a voice input is made.

If it is determined in S201 that a voice input has been made, the voice recognition unit 102 performs voice recognition processing (S202). At this time, as the recognition target vocabulary, the one selected from the recognition dictionary 104 by the recognition target vocabulary selection unit 103 is used.

In step S203, it is determined whether the voice command recognized in step S202 is an end command. If yes, the process ends there in response to the end instruction.

When it is determined to be no in S203, S2
04, it is determined whether the voice command recognized in S202 is an invalidation instruction. A normal voice command that is not an invalidation instruction has an action corresponding to the command registered in the recognition dictionary 104, and the voice command of the invalidation instruction is “invalid”, “cancel”, or “different”.
It is a command that must be estimated by the voice command performed immediately before, for the voice command in an arbitrary word such as "," the action instructing the invalidation of the voice command performed immediately before.

When the input voice command is a normal voice command (when it is judged as no in S204), S2
Go to 05.

In S205, the voice command is registered in the command recording unit 107 as an execution command, and the action stored in the recognition dictionary 104 corresponding to the voice command is sent to the action issuing unit 108.

In the next S206, the action issuing unit 10
Issue a voice command action from 8 to execute the command.

In the next step S207, the recognition target vocabulary selection unit 1
The recognition target vocabulary stored in 03 is updated as the command is executed, and the process returns to S201 to wait for the next input.

When it is determined in S204 that the command is an invalidation command (in the case of yes), the invalidation action estimation unit 105
In the command recording unit 107, the invalidation action for invalidating the command executed immediately before is estimated, the action is sent to the action issuing unit 108, and the command corresponding to the action is issued to the command recording unit. Register at 106. In S209, action issuing unit 10
Issue an invalidation action from 8 to execute invalidation.

In next step S210, the recognition target vocabulary selection unit 103
The recognition target vocabulary stored in is updated with the execution of the command. Then, in S211, the execution command registered in the command recording unit 107 in S205 is sent to the recognition target vocabulary selection unit 103, and the command is controlled to be excluded from the recognition target in the next recognition. It returns to S201 and waits for the next input.

3 to 5 show examples of invalidation in the embodiment of the present invention. FIG. 3 shows an example of invalidation of erroneous recognition when a menu is selected. When voice input is made as "Tochigi prefecture" in 1), it is erroneously recognized as "Miyagi prefecture" and is switched to the display of the city in Miyagi prefecture of the submenu. 2) If you input "No", you will be returned to the original menu and the display of the incorrectly recognized item "Miyagi Prefecture" will be corrected so that you cannot select it.
If you input "Tochigi Prefecture" again in 3), the submenu will change to display the cities in Tochigi Prefecture.

FIG. 4 shows an example of whether false recognition is invalid in the window operation. When "Reduce" is input by voice in 1), it is mistakenly recognized as "Delete", and the window is closed and deleted. If you input "No" in 2),
Open the file displayed in the window that was closed earlier with the same application and reproduce the previous state.
If you input "Reduce" again in 3), the window is miniaturized this time.

FIG. 5 shows an example of invalidation of graphic editing on drawing software. When voice input is made to "to the right" in 1), the triangle is erased because it is erroneously recognized as "cut". If you input "No" in 2), you can paste the erased triangle and reproduce the previous state. When the voice is input to "to the right" again in 3), the triangle is moved to the right this time.

In this embodiment, the action for invalidation is estimated from the command executed immediately before, but the action for executing the voice command is registered.
The user may register the dictionary as a voice command invalidating action.

Further, in the present embodiment, the voice command immediately before being invalidated is removed from the next recognition target vocabulary. However, this is returned to the original and the cancellation cancel voice command to be added as the recognition target vocabulary is added. You may prepare it.

Further, in the present embodiment, it is possible to provide a mode in which even the immediately preceding invalidated voice command is not excluded from the vocabulary to be recognized next time, and the mode can be switched by the user's designation.

Further, in the present embodiment, although the erroneously recognized voice command is returned to the state before execution and a new input is accepted, after returning to the state before execution, the immediately previous recognition result is A function may be provided that presents the next candidate to the user and allows the user to select from the next candidate and below.

Further, in the present embodiment, a new action is issued so as to restore the result of the malfunction, but it takes time to restore the original state, or the original state is restored. In the case of a voice command that causes an operation that is not guaranteed, a method of confirming the user before issuing the action may be used together.

Further, in the present embodiment, the invalidation instruction is given by a voice command such as "different", but this may be input by a mouse or a keyboard.

[0033]

As described above, according to the present invention,
The burden on the user is alleviated by repairing the change caused by the erroneous recognition and caused by the erroneous operation by a simple operation and by preventing the erroneous recognition from being repeated.

[Brief description of drawings]

FIG. 1 is a block configuration diagram of an embodiment of a voice processing device according to the present invention.

FIG. 2 is a flowchart of voice processing according to the embodiment.

FIG. 3 is a diagram showing an example of invalidating menu selection in the embodiment.

FIG. 4 is a diagram showing an example of invalidating a window operation in the embodiment.

FIG. 5 is a diagram showing an example of invalidation of graphic editing in the embodiment.

Claims (10)

[Claims] 1. A recognition means for recognizing inputted voice information, an information storage means for storing the information recognized by the recognition means, and a judgment as to whether or not the information recognized by the recognition means is an invalidation instruction. And a control means for controlling to execute processing for invalidating the processing according to the information stored in the information storage means when the determination means determines that the instruction is an invalidation instruction. An audio processing device having. 2. The voice processing apparatus according to claim 1, wherein the information stored in the information storage means is a command for executing processing. 3. The voice processing apparatus according to claim 1, wherein the information of the invalidated process is excluded from the target of the next voice recognition. 4. Recognizing the input voice information, determining whether or not the recognized information is invalidation instruction information, and when the determination determines that it is an invalidation instruction,
A voice processing method comprising controlling to execute a process of invalidating a process executed immediately before the voice information is input. 5. The voice processing method according to claim 4, wherein the information obtained by the recognition is a command for executing processing. 6. The voice processing method according to claim 4, wherein the information corresponding to the invalidated process is excluded from the target of the next voice recognition. 7. An input means for inputting voice information, a word dictionary for storing information as voice commands such as reading and notation of words, voice features and actions to be executed, and a vocabulary to be used for the current voice recognition. Selecting means, a voice recognizing means for recognizing an input voice by referring to a word dictionary for the selected recognition target vocabulary, a command determining means for determining the recognized voice as a voice command, and the determining A voice processing device that issues an action of a voice command that has been executed, a means that records the voice command executed in a straight line, a means that inputs an instruction to invalidate the execution result, and the execution of the voice command that was performed immediately before. A voice processing characterized by comprising means for estimating an action for invalidating the result, and means for issuing the estimated action. Apparatus. 8. The voice processing device, comprising means for excluding a voice command immediately canceled immediately thereafter from the current recognition target vocabulary. 9. Inputting voice information, referring to a word dictionary for a recognition target vocabulary whose vocabulary has been selected as a current voice recognition target, recognizing the input voice, and recognizing the recognized voice as a voice command. In the voice processing method for issuing the action of the determined voice command, the voice command executed immediately before is recorded, and the voice command executed immediately before is input in response to the input of the instruction to invalidate the execution result. A speech processing method, which estimates an action that invalidates the execution result of and executes the action. 10. A voice processing method, characterized in that a voice command immediately canceled is excluded from the current recognition target vocabulary.