JP2000293193A - Device and method for voice inputting and storage medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve the operating environment during a voice input processing by providing a recognizing means, which recognizes inputted voice, and an approving means which approves the reception of a next input while a voice recognition result is displayed without being confirmed when the degree of correctness of the voice recognition result of the recognizing means is low. SOLUTION: Voice signals inputted from a microphone 2 of a headset are taken into a main body through a connector 4 and amplified by an amplifier 13. Amplified voice signals are converted into digital signals by an AD converter 14 and transmitted to a control section 15. The section 15 executes programs and conducts voice recognition by using a voice recognition program held in a memory 17, a voice recognition registration dictionary, works to be used in the processes and voice data and controls a display section 6. When the degree of correctness of a voice recognition result is low, the result is displayed without being confirmed and an approval is given to a next input by automatically moving a cursor to a next input position (a next figure).

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice input technology for inputting various kinds of information by voice.

[0002]

2. Description of the Related Art Conventionally, in a handy terminal used for meter reading of a water usage meter, a voice input function of inputting information by voice using a voice recognition function has been realized.

In this type of electronic device, when the accuracy of the speech recognition result is low and the accuracy is uncertain, a plurality of candidates are displayed and selected from the candidates, or the candidates are input again by re-inputting. And then proceed to the next input.

[0004]

However, it is easy to perform an input confirmation operation on the display screen, but when a state where the display surface cannot be viewed or continuous input is desired, a screen for correction is required. Must be checked each time error recognition with low accuracy of the speech recognition result occurs, which is inconvenient.

The present invention has been made under such a background, and an object of the present invention is to improve an operation environment during voice input processing.

[0006]

Means for Solving the Problems (and Action) In order to solve the above problems, the present invention provides a recognition means for recognizing an input voice, and a method for recognizing a case where the accuracy of the voice recognition result by the recognition means is low. And permission means for permitting the next input while the voice recognition result is displayed in an undetermined state.

Further, the present invention provides a recognition step for recognizing an input voice, and, if the accuracy of the voice recognition result in the recognition step is low, the next input is displayed while the voice recognition result is displayed in an undetermined state. And a permission step of permitting

Further, the present invention is a storage medium storing a program for voice input processing, wherein the program comprises:
A recognition routine for recognizing the input voice; and a permission routine for permitting the next input while displaying the voice recognition result in an undetermined state when the accuracy of the voice recognition result by the recognition routine is low. .

Further, according to the present invention, when the accuracy of the speech recognition result by the recognition means is low, the corresponding input speech data is stored in the storage means / step / routine; Reproduction means, steps, and routines for reproducing the input audio data.

Further, the present invention has a warning means, a process, and a routine for warning, when the accuracy of the speech recognition result by the recognition means, the process, and the routine is low, more than a specified number of times. ing.

Further, the present invention has a moving means, a step, and a routine for automatically moving a cursor to the position of the unconfirmed speech recognition result at the time of shifting to the input confirmation processing.

Further, the present invention has a second permission means / step / routine for permitting a process of deciding an arbitrary unconfirmed position regardless of the input order at the time of the input deciding process.

Further, the present invention is applied to a handy terminal used for meter reading of a water usage meter or the like, wherein the voice input device (program for voice input processing) is applied.

[0014]

Next, embodiments of the present invention will be described with reference to the drawings.

FIG. 1 is an external view of a handy terminal for meter reading of a water usage meter to which the present invention is applied. FIG.
In FIG. 1, reference numeral 1 denotes a headset, which includes a microphone 2 and a speaker 3. Reference numeral 4 denotes a connector, through which the headset 1 can be detachably connected to the handy terminal 5. Reference numeral 6 denotes a display unit, and FIG. 1 shows, as an example, a screen on which an application program for meter reading is operating. The following description will focus on this meter reading program. Reference numeral 7 denotes a voice key for voice input, 8 denotes a cursor key, 9 denotes a voice reproduction key, 10 denotes a correction key, and 11 denotes an ENT key. Reference numeral 12 denotes a voice input area displayed on the display screen.

FIG. 2 is a block diagram showing a schematic electrical configuration of the handy terminal 5. As shown in FIG. In FIG.
Is a control unit for controlling the execution of a program, voice recognition, and the display unit 6, and is constituted by a CPU and the like. An audio signal input from the microphone 2 of the headset 1 is taken into the main body via the connector 4 and amplified by the amplifier 13. The amplified audio signal is digitized by the AD converter 14 and sent to the control unit 15. 16 is AD
These are control and data lines for controlling the transmission / reception of a conversion start command and data to the converter 14 and sending the converted data to the control unit 15. Reference numeral 17 denotes a memory, which is used for storing application programs, a voice recognition program, a registration dictionary for voice recognition, work used for work, and voice data.
It is composed of AM and the like.

An input unit 18 corresponds to the various keys shown in FIG. Reference numeral 20 denotes a DA converter, which converts digital data corresponding to voice transmitted from the control unit 15 via the control and data line 19 into analog. D
The analog signal converted by the A-converter 20 is amplified by the amplifier 21 and then output to the speaker 3 of the headset 1 via the connector 4 and can be heard as sound.

Next, the speech recognition processing will be described with reference to the flowchart of FIG.

When the open speech recognition process is started in step S1, sampling of data input from the microphone 2 via the AD converter 14 is started (step S1).
2). Then, it is determined whether or not voice input is performed by determining whether or not the level of the sampled data is equal to or more than a predetermined value (step S3). As a result, if the voice input has been performed, the process proceeds to the voice recognition program in step S4.

In the subroutine of the speech recognition program,
Speech data is sampled periodically as follows. That is, the sampled data (AD conversion value) is temporarily stored in the work area of the memory 17 (step S5), and the data is passed to the voice recognition program (step S5). Then, it is checked whether or not a certain period of silence has occurred in order to detect the end of the voice input (step S7). As a result, if no silence period longer than a predetermined time has occurred, the process returns to the voice recognition program in step S4. On the other hand, when a silence period of a predetermined time or more has occurred and a silence period (end of voice input) is determined, periodic sampling is stopped (step S8), and the end of sampling is notified to the recognition processing program. (Step S9), the process returns to the voice recognition program of Step S4.

In the speech recognition program, upon receiving the stop notification of the sampling, the speech recognition processing is performed based on the sampling data up to that time. The processing result of the voice recognition processing is shown in a format as shown in FIG. That is,
Flag 22 indicating normal recognition / error recognition shown in FIG.
For each of the first to fifth candidates shown in 23, the processing result of the voice recognition processing is indicated by items of number 24 of N1 to N5 and distance 25 from the sampling data of d1 to d5.

Here, the candidate number corresponds to a number in a dictionary registered in the memory 17 in advance, and a number, word or collocation corresponding to the number can be subtracted from the dictionary. The distance is a numerical value indicating “accuracy” of the recognition result, and the smaller the distance, the more accurate the speech recognition.

When the speech recognition processing in step S4 is completed, N1 as a first candidate is set as a recognition result (step S10), and it is checked whether or not the recognition is normal based on the flag 22 (step S11). As a result, in the case of error recognition, the flow proceeds to error processing in step S15. In the case of normal recognition, the distance d1 of the first candidate is a predetermined distance ds
It is checked whether it is smaller than (Step S12).
This is to check whether the first candidate is “accurate” with respect to a certain standard. If d1 is greater than ds, that is, if it is determined to be "inaccurate," step S15
Proceed to subsequent error processing.

On the other hand, when d1 is smaller than ds, that is, when it is determined to be "accurate", the process proceeds to step S13, where the difference between the distance between the first candidate and the second candidate is larger than the predetermined distance difference dt. It is determined whether or not. First candidate and second
A large difference between the candidates means that the accuracy of the first candidate is high, and a small difference between the first candidate and the second candidate means that the first and second candidates are similar and difficult to identify. Means that the accuracy is low. Therefore, if it is determined in step S13 that the difference between d2 and d1 is equal to or smaller than dt, the process proceeds to error processing in and after step S15. On the other hand, when it is determined in step S13 that the difference between d2 and d1 is larger than dt, normal recognition is performed, and the first candidate N1 is determined as a speech recognition result (step S14).

In the error processing, first, the flag 22 shown in FIG.
Is set (step S15), and an error number is set (step S16). This error number is a number assigned to identify the error recognition that has occurred this time, such as voice storage data. Next, the audio data temporarily stored in the work area in step S5 is stored in A
DPCM (Adaptive Differentia)
The data is compressed by a compression method such as 1 Pulse Code Modulation and stored again in the audio data area (step S17). Then, an error index is set (step S18), and the process ends (step S18).
19). The error index is an area for storing information necessary for reproducing audio data relating to the error, and as shown in FIG.
This is an area for storing a start address at which voice data is stored, the number of sampled data, and numbers N1 to N5 of first to fifth candidates as recognition results.

Next, a series of input processing of the meter reading program described above will be described with reference to the flowchart of FIG. This meter reading program is a program for inputting a six-digit meter number, that is, six voices, into the voice input area 12 of FIG.

When input processing is started in step S20, a screen necessary for input is first displayed (step S2).
1) It is checked whether the voice key 7 has been pressed (step S22). Voice input is valid when the voice key 7 is pressed. If the voice key 7 has been pressed, the process proceeds to voice input processing, where the number of inputs n is initialized to "1" (step S23), and the number of errors En is initialized to "0" (step S24). Next, it is determined whether or not the voice key 7 has been pressed (step S25). If the voice key 7 has been released, it is considered that the input has been interrupted, and the input processing is terminated in step S38. I do.

If the voice key 7 is still pressed,
A voice recognition process is performed (step S26). This voice recognition process is performed according to the flowchart shown in FIG. When the voice recognition processing is completed, the recognition result is returned in the format shown in FIG. Next, N1 which is the first candidate of the speech recognition result is set in the data D (n) (step S27). Here, n means 6-digit number 1-6. Then, it is checked whether or not the speech recognition result is normal recognition based on the flag 22 in the data indicating the speech recognition result (step S28). It is displayed at the display position corresponding to the number n (step S29). And
Input status flag F indicating normal recognition or error recognition
The normal flag is set to (n) (step S30), and the process proceeds to step S36.

If it is determined in step S28 that the speech recognition result is error recognition, the number corresponding to the N1 number is used to notify the user of the fact (that is, the speech recognition result is incorrect). The displayed word is highlighted at a display position corresponding to the input number n (step S31). Then, the returned error number is input to the input state flag F.
(N) (step S32), and the number of errors En
Is counted up (step S33).

Next, it is checked whether the number of errors En exceeds a predetermined number of times Ex (step S3).
4) If not exceeded, the process proceeds directly to step S36, and if exceeded, a buzzer is sounded to warn that the number of errors is large (step S35), and the process proceeds to step S36. If the error count En is cleared upon normal recognition without generating a warning sound at the number of errors as described above, the error recognition is continuously performed a certain number of times Ex.
A warning sound may be emitted when the number exceeds the threshold.

In step S36, the number of inputs n is counted up. Then, it is determined whether or not the number of inputs n exceeds “6” which is the number of digits of the input area 12 of the display screen (step S37), and if not, the process returns to step S25, whereby the input to the next digit is performed. It can be so. This means that the nth input is completed without discrimination between normal recognition and error recognition, and the next input can be performed. on the other hand,
If the number n of inputs exceeds “6” which is the number of digits of the input area 12 on the display screen, the input processing ends in step S38.

Next, a transition example of the input screen corresponding to the processing described in the flowchart of FIG. 7 will be described with reference to FIG. A six-digit voice input area 12 is displayed on the display unit 6, and the display state of the voice input area 12 changes, for example, as follows. First, in step S26, a cursor represented by a double line is displayed at the first digit, indicating that the input is the first digit. In this state, when the voice key 7 is pressed and "Ichi" is input by voice using the microphone 2 (step S27),
"1" is displayed in the first digit as the voice recognition result (step S28), and the cursor moves to the next second digit. Subsequently, when "San" is voice-inputted in step S29, "3" becomes 2
It is displayed at the digit (step S30), and the cursor moves to the next 3
Move to the digit.

Next, in step S31, "one" is voice-inputted, but in step S32, "7" is displayed in the third digit, which is highlighted because an error recognition has occurred as a voice recognition result. However, even if an error occurs, the cursor moves to the next fourth digit and the next digit can be input. Subsequently, by voice inputting "bee" in step S33, "8" is normally displayed in the fourth digit in step S34, and the cursor is moved to the next fifth digit. In step S35, "2" is displayed by voice input of "knee", and in step S36, the cursor is moved to the next sixth digit. However, this "2" is correct as a speech recognition result, but the accuracy of recognition is low, so that it is determined to be error recognition and is displayed in reverse video. Next, in step S37, "ku" is input by voice, and "9" is normally displayed in the sixth digit in step S38.

Next, a specific example of the error processing in steps S33, S34 and S35 in the flowchart of FIG. 7 will be described with reference to FIG. Assuming that the count number Ex at which the error warning buzzer sounds is set to, for example, "2", the error recognition has already occurred twice in the allowable limit in step S39. In this state, it is assumed that a voice input of "Knee" is input in step S40. In this case, “2” is correctly displayed as the recognition result, but the recognition accuracy is low, so that it is determined that an error has been recognized and the display is highlighted, and the number of errors exceeds Ex = 2 and becomes “3”. So, the buzzer sounds "pip" as a warning sound. This is to warn the user of the possibility that an error has occurred due to, for example, a shift in the position of the microphone when the number of errors is large. However, the cursor moves to the next digit and you are not forced to interrupt your input.

Next, the correction / determination operation after the completion of the voice input will be described with reference to FIG. Upon completion of the six-digit voice input, the correction screen shown in step S41 is displayed. On this correction screen, an "all re-input" button is displayed. The undetermined part (error recognition part) remains in reverse display. 2
The cursor represented by the double line can be moved with the cursor key 8. If "all re-input" is selected here, all the speech recognition results are initialized and the input processing is redone.
When the cursor is moved to an undetermined "7" position and the correction key 10 is pressed, whether the input state flag F (3) (n = 3 because of the third digit) is recognized as normal or error is recognized.
At the time of error recognition, the stored error information is read from the error number and error index in FIG.
42 individual correction screens are displayed.

When the correction key 10 is pressed,
If the cursor is automatically moved to a digit related to error recognition, usability is improved. Also, the correction order does not necessarily have to be performed in the input order, and the correction amount can be corrected in the order closer to the current cursor position, so that the amount of movement of the cursor can be reduced.

On the individual correction surface in step S42, the first to fifth candidates corresponding to the undetermined "7" are displayed. Here, when the audio reproduction key 9 is pressed, the compressed and stored audio data is restored, and the input audio corresponding to the undetermined “7” is reproduced from the speaker 3 via the DA converter 20. Here, the user may listen to the voice input by himself, check and select from the candidates, and select "1" as a result of hearing the reproduced voice in step S43, and confirm with the ENT key 11. When confirmed, the data D (3) is set to "1", the input state flag F (3) is changed to normal recognition, the error index is cleared, and the display is switched from inverted display to non-inverted display as shown in step S44. Indicates that the error has been cleared.

In the state of the correction screen instead of the individual correction screen, EN
Pressing the T key 11 means that all confirmations have been made. If the ENT key 11 is pressed down in step S44 while the fifth digit "2" is not yet confirmed, a confirmation screen shown in step S45 is displayed. You. Here, if it is sufficient to keep “2”, the ENT key 11 is pressed to execute all determinations, and the correction is ended. Correction key 1 if "2" is wrong
Pressing 0 returns to the correction screen and performs the individual correction described above.

Next, another modification will be described with reference to FIG. Step S46 is a correction screen, but there is no undetermined peak. Here, the cursor is moved to the sixth digit “9” and the correction key 10 is pressed. In other words, this is a correction example in the case where no error recognition has occurred but the speech recognition result is wrong, or in which a mistake has been made. In this case,
Since the input state flag F (6) has been completed normally and the input voice information has not been stored, re-input by voice is waited as shown in step S47. At this point, since the original "9" has not been erased, pressing the correction key 10 returns to the original state.

Here, if the voice key 7 is depressed and "Ichi" is input by voice, the screen shifts to the step S49. Here, in order to ensure input, the first to fifth candidates are displayed and selected with the cursor to confirm. That is, when the first candidate “1” is selected and the ENT key 11 is pressed, the sixth digit is determined to be “1”, and the data D (6)
Is set to "1", and normal recognition is set to F (6). When the ENT key 11 is pressed in this state, that is, in a state where there is no undetermined digit, a voice input is made as shown in step S51.

As described above, the next input can be performed without being determined, so that the operation is not interrupted by error recognition. Also, when making corrections collectively later,
It can be confirmed using the input voice, and correction is easy. Furthermore, since a warning is issued when an error is recognized more than a certain number of times, a warning is not issued every time an error is recognized.
An abnormal input environment can be avoided.

Since individual corrections do not necessarily have to be made in the order of input, corrections can be made in the order closer to the current cursor position. Is displayed, the usability is improved, for example, the cursor can be automatically moved to the error recognition position.

The present invention is not limited to the above-described embodiment, but may be applied to any electronic device other than a handy terminal for meter reading of a water supply or electricity meter as long as it has a voice input function. It is also possible.

[0044]

As described above, according to the present invention,
A recognition unit for recognizing the input voice; and a permission unit for permitting the next input while the voice recognition result is displayed in an undetermined state when the accuracy of the voice recognition result by the recognition unit is low. When the user cannot view the display surface or wants to perform continuous input, it is no longer necessary to check the screen for correction every time error recognition with low accuracy of the voice recognition result occurs. The operating environment during processing is improved and usability is improved.

[Brief description of the drawings]

FIG. 1 is an external view of a handy terminal to which the present invention is applied.

FIG. 2 is a block diagram showing an electrical schematic configuration of the handy terminal.

FIG. 3 is a flowchart illustrating a speech recognition process.

FIG. 4 is a flowchart continued from FIG. 3;

FIG. 5 is a conceptual diagram showing a format of recognition result information.

FIG. 6 is a conceptual diagram illustrating an error index.

FIG. 7 is a flowchart showing a voice input process.

FIG. 8 is a diagram showing an example of transition of a display screen at the time of voice input processing.

FIG. 9 is an explanatory diagram for explaining an example of error processing.

FIG. 10 is an explanatory diagram for explaining an example of a confirmation operation for voice recognition.

FIG. 11 is an explanatory diagram for explaining another example of a confirmation operation of voice recognition.

[Explanation of symbols]

 1: Headset 2: Microphone 3: Speaker 5: Handy terminal 6: Display unit 7: Voice key 8: Cursor key 9: Voice playback key 10: Correction key 11: ENT key 12: Voice input area 15: Control unit 17: Memory 18: Input unit 22: Flag indicating normal recognition / error recognition

Claims (18)

[Claims] 1. A recognition means for recognizing an input voice, and a permission for permitting a next input while displaying the voice recognition result in an undetermined state when the accuracy of the voice recognition result by the recognition means is low. A voice input device comprising means. 2. A storage unit for storing corresponding input voice data when the accuracy of the voice recognition result by the recognition unit is low; a reproducing unit for reproducing the input voice data stored by the storage unit; The voice input device according to claim 1, further comprising: 3. The apparatus according to claim 1, further comprising a warning unit for warning when the accuracy of the result of the voice recognition by the recognition unit is low more than a predetermined number of times. A voice input device as described. 4. The apparatus according to claim 1, further comprising a moving unit for automatically moving a cursor to a position of the unconfirmed speech recognition result when shifting to input confirmation processing. Voice input device. 5. The apparatus according to claim 1, further comprising a second permitting unit for permitting a process of deciding an arbitrary unconfirmed position regardless of an input order during the input deciding process. Voice input device. 6. The voice input device according to claim 1, wherein the voice input device is applied to a handy terminal used for meter reading of a water usage meter or the like. 7. A recognition step of recognizing an input voice, and a permission to permit a next input while displaying the voice recognition result in an undetermined state when the accuracy of the voice recognition result in the recognition step is low. A voice input method comprising a step. 8. A storing step of storing the corresponding input voice data when the accuracy of the voice recognition result in the recognition step is low; a reproducing step of reproducing the input voice data stored in the storing step; The voice input method according to claim 7, further comprising: 9. The method according to claim 7, further comprising a warning step of warning when the accuracy of the speech recognition result in the recognition step is low more than a predetermined number of times. The described voice input method. 10. The method according to claim 7, further comprising a moving step of automatically moving a cursor to a position of the unconfirmed speech recognition result when the process proceeds to the input confirming process. Voice input method. 11. The method according to claim 7, further comprising a second permitting step of permitting a process for deciding an arbitrary unconfirmed position irrespective of an input order at the time of the input deciding process. Voice input method. 12. The voice input method according to claim 7, wherein the voice input method is applied to a handy terminal used for meter reading of a water usage meter or the like. 13. A storage medium storing a voice input processing program, the program comprising: a recognition routine for recognizing an input voice; and a low accuracy of a voice recognition result by the recognition routine. A storage medium comprising a permission routine for permitting a next input while the voice recognition result is displayed in an undetermined state. 14. A storage routine for storing corresponding input voice data when the accuracy of the voice recognition result obtained by the recognition routine is low; a reproduction routine for reproducing the input voice data stored in the storage routine; 13. The storage medium according to claim 12, comprising: 15. The apparatus according to claim 12, further comprising a warning routine for warning when the accuracy of the speech recognition result obtained by said recognition routine is low more than a predetermined number of times. The storage medium according to the above. 16. A moving routine for automatically moving a cursor to a position of the unconfirmed speech recognition result when shifting to input confirmation processing.
16. The storage medium according to any one of 15). 17. A method according to claim 13, further comprising a second permission routine for permitting a process of deciding an arbitrary unconfirmed position irrespective of the input order at the time of the input deciding process. Storage medium. 18. The program for speech input processing,
3. The method according to claim 1, wherein the step is performed in a handy terminal used for meter reading of a water usage fee meter or the like.
8. The storage medium according to any one of 7.