JPH064264A - Voice input/output system - Google Patents

Abstract

PURPOSE:To prevent an erroneously recognized command from being executed by confirming contents of a recognized command in the interactive system by a response message before execution of this command. CONSTITUTION:After a voice key emulator is started, various functions of voice registration to register the contents of a voice recognition dictionary in a dictionary file 4, dictionary file load to write the registered contents of the dictionary file 4 in the memory of a personal computer, saving, recognition test, key code editing, etc., are executed. When the command name of a command to be executed by an operator is generated after depression of a recognition execution key on a keyboard 8 at an arbitrary time during execution of general purpose soft thereafter, its voice is inputted through a microphone 1 and is recognized. Before a key stroke string corresponding to the recognition result of this voice input is executed, the response message to prompt the operator to confirm this key stroke string is displayed with a voice.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice input / output system which has a function of recognizing a voice and a function of synthesizing a voice and is capable of deciding execution or rejection of a command of a keystroke sequence in a conversational format, The present invention relates to a voice input / output system that makes it difficult for an incorrect application program to be executed due to erroneous recognition of voice input.

[0002]

2. Description of the Related Art Generally, a command for controlling the operation of an electronic system such as a computer system is input by a hand-operated input means such as a keyboard and a mouse. To overcome obstacles, it has been attempted to be able to enter these commands by voice.

In an electronic system having a function of inputting a command by voice, in order to realize input of a key string similar to keyboard input by voice, a feature amount obtained by inputting voice and a standard registered in advance are used. A means for recognizing a voice input by comparing with a pattern, and a keystroke string sequence previously edited by the user for each pre-registered individual word pattern, and a corresponding key based on the previous recognition result. A key emulator means for transferring a stroke ring string to a keyboard buffer for keyboard input, and an interrupt signal generated in a running application program included in the key emulator means by a predetermined key stroke, and the character string And a means for editing the voice input system is provided.

Such a voice input system recognizes voice at a word level or a sentence level, and is useful for enhancing operability by replacing a complicated keystroke for command input such as an application program having a word processor function with a voice. It is being attempted to use it for.

It is necessary for the above-mentioned key stimulator means to run in parallel with the running application program as in the case of controlling the keyboard, but such control is not disclosed in, for example, Japanese Patent Publication No. 3-206521. Can be realized by adopting the method described above.

[0006] When a command is input by voice like this,
It is necessary to prevent the content of the command from being erroneously recognized and execute the application program according to the erroneously recognized command.

In order to prevent the execution of the application program in accordance with the erroneously recognized command, in the conventional electronic system having the voice input function, a response message prompting confirmation of execution of the application program corresponding to the recognized command is displayed. Image on the display,
Only when there is a confirmation input given within a predetermined time after this display, or when there is no confirmation or rejection input within a predetermined time after this display. It is configured to execute an application program.

The response message displayed in response to the voice input is registered in advance in the storage means, or the voice data generated at the time of dictionary registration is used to make a character string registered in advance at the time of response. It is made according to.

[0009]

As described above, in the conventional voice input device, since the response message having the contents registered in advance is displayed, the displayed contents are unique and the user can know the accuracy of the recognition result. I can't.

Further, since the application program is executed according to the command corresponding to the recognition result of the voice input, when the command program is erroneously recognized and the wrong application program is executed, the erroneously executed application program is executed. Restoring a previous state is very cumbersome, and it may not be possible to restore the state, for example if an application program that erases data is accidentally executed.

Particularly, in the voice key stimulator, unlike the keyboard operation, a plurality of command operations can be performed by one operation.
Since it is often assigned to one voice command, repairing at the time of erroneous recognition often requires more effort than when a command is erroneously made by the keyboard.

Further, particularly when the recognition result is not displayed by the response message, it is also unknown which application program will be executed, and the correctness of the executed application program is recognized from the execution result of the application program to repair it. There is no choice but to plan.

In view of the above circumstances, it is an object of the present invention to provide a voice input / output system that makes it difficult for an erroneous recognition of voice input to cause an incorrect application program to be executed.

[0014]

According to the present invention, an input means for inputting a control command by manual operation, a voice input means for inputting a control command by voice, and a feature amount obtained by inputting voice are registered in advance. A means for recognizing a voice input by comparing with a standard pattern that has been stored, and a keystroke string sequence that has been previously edited by the user for each preregistered individual word pattern, based on the previous recognition result. A key emulator means for transferring a corresponding keystroke ring sequence to a keyboard buffer for keyboard input, and an interrupt signal generated on a running application program included in the key emulator means by a predetermined keystroke. , In a voice input / output system including means for editing the character string, To formed, it has taken the following means.

That is, the input voice is pre-encoded and compressed, recorded at the word or sentence level, and stored as a response message or each phoneme data is provided at the voice segment level, and a given character string is provided. And a means for generating a response message according to the above, and when executing the keystroke sequence corresponding to the recognition result in response to the voice input, the keystroke sequence is designated in advance, and confirmation of execution of the keystroke sequence is prompted. The response message and the addition to this,
It is characterized by comprising a display means for responding with a response message prompting confirmation of execution of the keystroke sequence other than the keystroke sequence before executing the command of the keystroke sequence.

[0016]

In the present invention, a response message prompting confirmation of the keystroke sequence before execution of the keystroke sequence corresponding to the recognition result of the voice input, and a response message added thereto,
Since a response message prompting confirmation of execution other than the keystroke sequence is displayed, the contents of the keystroke sequence to be executed from now on will be clearly distinguished from the others, and the recognition result of voice input will be displayed. The operator can clearly recognize whether or not is correct, and if the recognition result is incorrect based on this clear recognition, the operator must take a measure to reject the execution of the misrecognized keystroke sequence. Is prompted.

[0017]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A voice input / output system according to an embodiment of the present invention will be described below in detail with reference to the drawings.

As shown in the block diagram of FIG. 1, a voice input / output system according to an embodiment of the present invention is used for inputting a command to a personal computer 9, and its voice key stimulator is an operator. A microphone 1 for inputting a voice given by, and a pre-processing unit 2 for removing noise from the sound collected by the microphone 1.
The pre-processing unit 2 includes a feature extraction unit 3 that extracts an unknown parameter of speech from which noise has been removed, and a dictionary unit file 4.

In the dictionary file 4, the voice recognition dictionary unit 41 for registering the feature of the voice of the operator who has been patterned is registered.
An execution key string dictionary unit 42 for registering an execution key string corresponding to each voice pattern registered in the voice recognition dictionary unit 41, each voice pattern registered in the voice recognition dictionary unit 41, and an execution key corresponding thereto There are provided a response message data section 43 in which response message data corresponding to a string is registered, and a guidance message data section 44 in which guidance message data for differently expressing the response message is registered.

The voice key stimulator also includes
A recognition processing unit 5 is provided that compares the voice feature of the operator extracted by the feature extraction unit 3 with each voice pattern registered in the voice recognition dictionary unit 41 and recognizes a similar voice pattern.

As a speech pattern recognition method, dynamic programming (DP: Dynamic Programming), frequency components at each time of TSP (Time-Cpectrum-Pattern) with time on the vertical axis and frequency on the horizontal axis are used. Binarized BTSP (Bynar
Although various methods such as the BTSP method using y-Time-Cpectrum-Pattern) can be adopted, the BTSP method disclosed in Japanese Patent Publication No. 3-42480 is adopted here.

As a voice recognition method, there are a method of performing single syllable recognition corresponding to each key of the keyboard, a method of performing word recognition so as to correspond to a keystroke in a word unit, etc. Since monosyllabic recognition in which recognition is generated by dividing keystrokes is natural, speech recognition is performed using word recognition that is generally used.

Therefore, in the voice recognition dictionary unit 41, for example, "centering", which is a pattern of operator voice features, is registered. Also, the keystroke string corresponding to the voice pattern "centering" registered in the voice recognition dictionary unit 41 is "CENTE".
RING ", and the execution key string in which" ESC "indicating the command is added to the character string of" p / s "set as the command of the corresponding application is the execution key string as" ESC p / f ". It is registered in the dictionary unit 42. Further, in the response message data section 43, data “centering” corresponding to “centering” registered in the voice recognition dictionary section 41 is registered.

The voice key stimulator also includes
When the recognition processing unit 5 recognizes a voice, the result processing unit 6 that reads the execution key string corresponding to the recognized voice pattern from the execution key string dictionary unit 42 and the execution that is read by the result processing unit 6 A key string output unit 7 for outputting a key string is provided.

The execution key string output from the key string output unit 7 is input to the input buffer of the keyboard 8, and a command similar to the key command corresponding to the key command corresponding to this execution key string is input from the keyboard 8 to the personal computer 9. Is output to the personal computer 9 according to this command.

Further, this voice key stimulator has
Before the personal computer 9 executes an application program by the execution key string from the key string output unit 7, a message display is displayed to form a response message in response to a voice input in order to prompt confirmation of the execution of the application program. Means 10 are provided.

The message display means 10 prompts confirmation of execution of response message data corresponding to the voice pattern recognized by the recognition processing section 5 from the response message data section 43 and the guidance message data section 44. A response that reads the three strings "result is", "is.", And "may I execute", and synthesizes the string "result is centering. A data synthesizing unit 11, a data decoding unit 12 for decoding the synthesized response message into a digital voice signal corresponding to the contents, and a D / A conversion unit 13 for converting the decoded digital voice signal into an analog voice signal.
And a speaker 14 that inputs the analog voice signal and outputs a voice corresponding to the content.

Next, the operation of this voice input / output system will be specifically described with reference to FIG.

First, after the voice key emulator is activated by commercially available general-purpose software used in the personal computer 9, the voice registration for registering the contents of the voice recognition dictionary in the dictionary file 4 and the registration contents of the dictionary file 4 for the personal computer are performed. Key code for loading a dictionary file to be written in the memory of the personal computer 9, writing the registered content written in the memory of the personal computer 9 to the dictionary file 4, saving the dictionary file 4, a recognition test, and editing the dictionary registered in the dictionary file 4. Various functions of the voice key emulator such as editing are executed.

After that, the voice utility is kept resident in the main memory of the personal computer 9, the program of the general-purpose software is returned, and the recognition execution key provided on the keyboard 8 is pressed at any time during the execution of the general-purpose software. When the operator generates a command name to be executed, the voice is input and recognized through the microphone 1, and the command registered in the key emulator software can be executed by voice instead of keyboard input.

[0031] For example, a response message prompting confirmation of voice input of the above command name and execution of the recognized command or the execution of a command other than the recognized command after the recognition is made, "The result is centering. Is it okay to run? "

After the output of this response message, the program returns to the general-purpose software, and the operator responds to this response message within a predetermined time by keyboard input or voice input (for example, voice input of "Yes" or "execute"). Confirm whether to execute the command that was input by voice, whether to execute another command by keyboard input or voice input (for example, input of voice such as "No", "Cancel", or "Reject"), or within a predetermined time If the operator does not make such confirmation, the confirmation of execution of the command input by voice is assumed. Then, the command according to the confirmation result is executed in the microcomputer 9.

As described above, a response message prompting confirmation of the keystroke sequence before executing the keystroke sequence corresponding to the recognition result of the voice input, and a confirmation message added to the response message to confirm the execution of other than the keystroke sequence. Since the prompting response message is displayed by voice, the contents of the keystroke sequence to be executed from now on will be clearly distinguished from the others, and the operator will be asked whether or not the recognition result of voice input is correct. If the recognition result is clear, and if the recognition result is incorrect based on this clear recognition, the operator takes steps to reject the execution of the misrecognized keystroke sequence, thereby executing the command due to misrecognition. Is prevented.

In the voice input / output system according to another embodiment of the present invention shown in FIG. 3, the similarity between the voice pattern input from the recognition result and the voice pattern registered in the voice recognition unit 41 is determined. The degree determination section 20 is additionally provided, and the guidance message data section 44 provides a guidance message when the degree of similarity is high, for example, “result is”,
Area 44a that registered high-level messages such as "Is it?" And "Is it okay to execute?", A guidance message when the similarity is medium, such as "Result is", "I think.", "Execution" Area 44b for registering a middle-level message "Is it okay?" And a guidance message when the degree of similarity is low, such as "Result is".
please check. An area 44c for registering a low order message "" is provided. Also, the message display means 1
0 is additionally provided with a response data switching unit 15 that switches the guidance message read by the response data synthesizing unit 11 according to the determination result of the similarity determining unit 20 according to the similarity.

In this voice input / output system, FIG.
As shown in the flowchart of FIG. 5, when a voice is input (step S1), the recognition process (step S2) is executed, and then the similarity determination unit 20 determines the similarity of the recognition result (step S3). .

Thereafter, the response data synthesizing unit 11 creates different response messages corresponding to the degree of similarity (steps S4 to S6). For example, if the similarity is low, a low-level response message "The result seems to be centering. Please check" is created (step S4), and if the similarity is medium, "I think the result is centering. A middle-level response message "Do you want to execute?" Is created (step S5), and if the recognition results have a high degree of similarity, a high-level response message "The result is centering. It is produced (step S6). Then, the produced response message is output from the speaker 14 (step S
7) The operator can audibly recognize the degree of execution confirmation.

Output of this response message (step S
7) After that, if the operator does not give an instruction by voice or a key within a predetermined time, the confirmation of execution is pretended and the command input by voice is executed (step S).
9).

In this voice input / output system, when the command name displayed by the response message is different from the operator's intention, the operator uses the microphone 1 by voice within a predetermined time after the response message is displayed. By giving the next instruction such as "next" or "next candidate" or by manually inputting the keyboard 8 or a mouse (not shown) (steps S8 and S10), the previously input voice is similar to the next. A response message in which the command name is changed to a command name of a recognition result having a low degree of similarity is produced again (step S11), and the response message is output again from the speaker (step S11).
7).

In this case, the modified response message is displayed with a message having a different expression from the first response message, for example, "Next candidate is ... Do you want to execute?" The operator can be made aware that the value has been changed.

The procedure of revision of the response message (step S8, step S10, step S11) and the output of the revised response message (step S7) are as follows.
Confirmation is performed without operator's voice input within a predetermined time after displaying the response message, or repeated until voice input or manual input is given to reject the command by the voice input of “stop” or “cancel”. (Step S8, Step S10, Step S12), and finally the command for which confirmation is assumed is executed as a voice input command (Step S9).

If a voice input or a manual input for rejecting the voice input command is given (step S8,
In step S10 and step S12), the command by the voice input is rejected (step S13).

Further, in this embodiment, by manually inputting a command different from the voice input instead of the voice input or the manual input for rejecting the recognized command, the command input by voice is rejected. It may be configured to execute a manually input command.

Since the other construction, operation and effect of this embodiment are the same as those of the above-mentioned first embodiment, their description will be omitted to avoid duplication.

In each of the above embodiments, the response message is displayed by the voice output from the speaker 14, but instead of or together with this, an image connected to the personal computer 9, for example. The response message may be displayed in characters on the screen of the display device.

[0045]

As described above, according to the present invention,
When an operator executes a command input by voice, the operator can confirm the contents of the command interactively by a response message before the recognized command is executed, and a measure to reject the execution of the incorrectly recognized command By doing so, it is possible to prevent the command that is erroneously recognized from being executed.

Further, in the present invention, the response message prompting confirmation of execution of the keystroke sequence corresponding to the recognition result is added with the response message prompting confirmation of execution of a sequence other than the keystroke sequence, and displayed by means of the hinge means. Therefore, the contents of the keystroke sequence that is about to be executed will be displayed clearly distinguished from other keystroke sequences, and the operator will clearly recognize whether the recognition result of the voice input is correct or not. Therefore, it is possible to more reliably prevent the command that is erroneously recognized from being executed.

In the present invention, there is provided means for varying the content of the response message stored in the means for storing the response message or the content of the response message generated by the means for generating the response message depending on the degree of similarity of the recognition result. In this case, it is possible to know the degree of confirmation that the operator should make from the displayed contents, so even if the voice is input in a state where it is difficult to recognize it, the operator needs to re-enter the command by taking appropriate confirmation measures. It is possible to improve the operability.

Further, in the present invention, the command of the keystroke sequence is executed only when the confirmation of the execution or the input of the rejection by the input means or the voice is not input within a predetermined fixed time after the response by the display means. With such a configuration, it is not necessary to perform a special confirmation operation when the voice input is correctly recognized, and the operability can be improved.

Further, in the present invention, the similarity of the recognition result is obtained especially when there is an input by the input means or an input by voice which points out an error in the recognition content within a predetermined time after the response by the display means. In the case where the display means is provided with a response message for prompting confirmation of the recognition result of the second candidate having a low recognition rate, the next candidate can be easily and easily input by pointing out an error in the recognition content. Can be easily selected and operability can be improved.

In addition, in the present invention, execution or rejection of the command of the displayed keystroke sequence is determined according to the content of the input means or voice input given within a predetermined time after the response by the display means. In this case, it is possible to cancel the command that was already recognized by voice input before executing the command that was already input by voice input without using manual input using the keyboard or mouse, so it is possible to prevent the execution of the command that was erroneously recognized by voice input. The operability can be further enhanced.

[Brief description of drawings]

FIG. 1 is a block diagram of a voice input / output system according to an embodiment of the present invention.

FIG. 2 is a flowchart showing a flow of operations of the voice input / output system according to the embodiment of the present invention.

FIG. 3 is a block diagram of a voice input / output system according to another embodiment of the present invention.

FIG. 4 is a flowchart showing a flow of operations of a voice input / output system according to another embodiment of the present invention.

[Explanation of symbols]

 1 Microphone 3 Feature Extraction Section 4 Dictionary File 5 Recognition Document Section 6 Result Processing Section 8 Keyboard 9 Personal Computer 10 Message Display Means 41 Voice Recognition Dictionary Section 42 Execution Key String Dictionary Section 43 Guidance Message Data Section

Claims (5)

[Claims] 1. An input means for inputting a control command by hand operation, a voice input means for inputting a control command by voice, and a feature amount obtained by inputting voice and a standard pattern registered in advance. And a means for recognizing voice input, and a keystroke string sequence previously edited by the user for each preregistered individual word pattern,
In a voice input / output system comprising a key emulator means for transferring a corresponding keystroke ring sequence to a keyboard buffer for keyboard input based on the recognition result, the input voice is encoded and compressed in advance, and a word is input. Alternatively, a means for recording at the sentence level and storing it as a response message or a means for generating a response message according to a given character string having each phoneme data at a voice segment level, and recognizing in response to the voice input When executing the keystroke sequence corresponding to the result, a response message which is designated in advance in the keystroke sequence and prompts confirmation of execution of the keystroke sequence, and a response message added to the response message and confirmation of execution other than the keystroke sequence. And a response message prompting the user to respond before executing the command of the response stroke sequence. Speech output system, characterized in that it comprises a. 2. The means for storing the response message, wherein the content of the response message stored in the means for storing the response message or the content of the response message generated by the means for generating the response message differs depending on the degree of similarity of the recognition result. The voice input / output system according to 1. 3. A means for executing a command of a keystroke sequence only when there is no confirmation of execution or rejection of input by the input means or voice within a predetermined time after the response by the display means. The voice input / output device according to claim 1. 4. A second candidate having a low degree of similarity in recognition result when there is an input by the input unit or an input by voice that points out an error in the recognition content within a predetermined time after the response by the display unit. The voice input / output device according to claim 1, further comprising: a unit that causes the display unit to respond with a response message prompting confirmation of the recognition result. 5. The execution or rejection of the command of the displayed keystroke sequence is determined according to the content of the input means or voice input given within a predetermined time after the response by the display means. Or the voice input device according to 4.