JP2002108390A - Speech recognition device and computer-readable recording medium - Google Patents

Abstract

(57) [Summary] [Problem] To reduce the omission of a first part and a last part of a voice section to be extracted, and to reduce a troublesomeness of making a user aware of a timing of pressing a switch and a timing of utterance. An audio input unit, a switch state detection unit for detecting an operation state of a switch, and an input control unit are provided. When the use of the voice input unit 1 is notified to the voice input unit 1 and the use of the voice signal is no longer necessary, the voice input unit 1 is notified of the end of use of the voice input unit 1 and the switch 3 is operated. When the state changes from OFF to ON, the audio input unit 1 is immediately notified of the start of the capture of the audio signal, and the switch 3 is turned ON to OFF.
When the state becomes the F state, the end of the capture of the audio signal is notified with a delay of a predetermined time.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice recognition device and a computer-readable recording medium which fetch a required voice signal in conjunction with a switch operation and recognize the fetched voice signal.

[0002]

2. Description of the Related Art At present, personal computers (hereinafter, referred to as personal computers).
When a predetermined operation is performed on a machine such as a personal computer, a method using a keyboard, a mouse, or the like is mainly used. However, when displaying a frequently viewed Web page on a personal computer, for example, it is necessary to start up a Web browser and select a Web page registered as a “favorite” or input a complicated URL. is there. Therefore, voice recognition which gives a command by directly inputting a voice and performs a predetermined operation has been attracting attention.

Conventionally, as this type of speech recognition apparatus, FIG.
As shown in the figure, an audio signal 11 obtained by digitizing an analog audio signal collected by a microphone with an A / D converter is taken in, and the audio power calculated from the audio signal 11 is given to an audio section cutout unit 12 to provide an audio section. Cut out
A feature vector of the cut-out speech section is extracted by the acoustic analysis unit 13 and provided to the matching unit 14. The matching unit 14 performs matching with a standard pattern registered in advance in the dictionary unit 15, so that There is a device that outputs a similar result as the recognition result 16.

In this case, a method for detecting a voice section is as follows.
For example, with respect to the audio power calculated from the audio signal, a start point at which the audio power is increased to a predetermined value or more is detected, and a temporary end point at which the audio power is equal to or less than a first threshold is detected, In some cases, the tentative end point is sequentially moved to a position of less than and the audio signal from the start point to the end point is cut out as an audio section in response to the determination of the final tentative end point as the end point.

However, if a voice section is detected while a voice signal is constantly being captured, an unnecessary word such as "um" or sudden surrounding noise (noise) may be included in the captured voice signal. However, these unnecessary words and noise sections are erroneously detected as speech sections, and recognition is performed.

Therefore, in practice, the user directly specifies a necessary voice section, such as a method of receiving a voice signal input while the switch is being pressed or a method of receiving a voice input for a certain time after pressing the switch once. By combining the method and the above-described detection of the voice section, erroneous detection of the voice section due to unnecessary words, ambient noise, and the like is prevented.

[0007]

By the way, in the above method, when the user presses the switch at least slightly after the start of the utterance, or when the user releases the switch at least a little before the end of the utterance of the user. However, there is a problem that the first part and the last part of the voice section to be extracted are missing.

As one method for solving this, in Japanese Patent Application Laid-Open No. Hei 8-185196, an input audio signal is always captured and stored, and stored over a wider range than a detection section designated by a switch operation. Only one voice section is extracted from the input voice signal and output.

However, when this method is used in a personal computer or the like, one application software occupies the voice input device, and even if another application software that requires voice input is operated, the voice input device is used. A problem arises that it cannot be performed. In addition, when this method is used in a portable information terminal or the like, which is one of the important factors of being able to use for a long time with low power consumption, the power consumption increases because the audio signal is always taken in, so that the power consumption increases. There is a problem that the available time is shortened.

The present invention has been made in view of such circumstances, and in a method in which a user directly specifies a necessary voice section using a switch, the first section or the last section of the voice section to be extracted is missing. And reduce the hassle of making the user aware of the timing of pressing the switch and the timing of the utterance as much as possible.Also, the power consumption can be reduced by operating the voice input unit only when necessary and capturing the voice signal. Yes, and one application software does not always occupy the voice input unit,
It is an object of the present invention to provide a voice recognition device capable of performing input control that can use a voice input unit from other application software, and a computer-readable recording medium storing such a processing program.

[0011]

A voice recognition device according to the present invention is a voice recognition device for capturing a voice signal in response to a switch operation for designating the capture of a voice signal and recognizing the captured voice signal. Unit, a switch state detection unit for detecting an operation state of the switch, and an input control unit, and the input control unit uses the voice input unit for the voice input unit when it is necessary to capture a voice signal. When the switch from the OFF state to the ON state is sent to the voice input unit, the voice input unit is notified immediately when the use of the audio signal is no longer necessary. A function shall be provided to notify the start of audio signal capture and notify the end of audio signal capture after a predetermined time when the switch changes from ON to OFF. Thus characterized.

According to the speech recognition apparatus of the present invention, it is possible to reduce the loss of the first part and the last part of the speech section to be extracted, and to make the user aware of the timing of pressing the switch and the timing of utterance. The annoyance can be reduced as much as possible. In addition, since the voice input unit is operated to capture the voice signal only when necessary for voice recognition, power consumption can be reduced.

[0013] In the voice recognition device of the present invention, the input control unit operates when the switch is turned from OFF to ON.
Immediately notifies the voice input unit of the use of the voice input unit and the start of capturing the audio signal, and switches from ON to OFF
A function may be provided for notifying the audio input unit of the end of capturing the audio signal and the end of use of the audio input unit after a predetermined time delay when the state is entered.

According to the present invention, ON / OF of the switch is performed.
Since the audio input unit is operated according to F to capture an audio signal, power consumption can be further reduced.

[0015] In the voice recognition device of the present invention, the input control unit operates when the switch is turned from OFF to ON.
Immediately notify the voice input unit of the start of voice signal capture, and enable voice input to the user after a preset time has elapsed since the switch was turned on. May be provided.

According to the present invention, the audio signal is fetched shortly before the notification (for example, display of an icon, etc.) indicating the state in which the voice can be input.
Even if the user speaks a little earlier than the notification indicating the state in which the voice can be input, it is possible to reduce the lack of the beginning voice.

In the voice recognition device of the present invention, the input control unit terminates the capture of the voice signal to the voice input unit after a preset time has elapsed since the switch was turned from ON to OFF. A function for notifying may be provided.

According to the present invention, the lack of voice at the end can be reduced even if the user does not notice the timing of releasing the switch.

In the speech recognition apparatus according to the present invention, the input control unit has a speech section detection function and, after the switch is turned from ON to OFF and the detection of the speech section has been completed, the input control unit controls the speech input unit. May be provided for notifying the end of the capture of the audio signal.

According to the present invention, since the voice input unit does not end capturing the voice signal during the voice section, the loss of voice in the last part of the voice section can be reduced.

The recording medium of the present invention captures an audio signal in response to a switch operation for designating the capture of an audio signal,
A computer-readable recording medium on which a processing program for recognizing a captured audio signal is recorded. Notifies the voice input unit of the end of use of the voice input unit when capture is no longer required, and starts capturing voice signals to the voice input unit immediately when the switch is turned on from OFF. And a processing program for notifying the end of the capture of the audio signal with a delay of a predetermined time when the switch changes from the ON state to the OFF state.

[0022]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing the configuration of an embodiment of the speech recognition apparatus of the present invention.

The speech recognition apparatus shown in FIG. 1 is mainly composed of a speech input unit 1, an input control unit 2, a switch 3, a switch state detection unit 4, a speech recognition unit 5, a display unit 6, and the like.

The audio input unit 1 digitizes an analog audio signal collected by the microphone 1A with an A / D converter 1B. The digitized voice signal is input to the input control unit 2 and the voice recognition unit 5.

The input control unit 2 controls the input of the audio signal by controlling the audio input unit 1 in accordance with the state of the switch 3 and the detection state of the audio section.

The switch 3 generates an ON / OFF signal by a user's operation of the switch. ON and OFF signals are
The signal is input to the switch state detector 4.

The switch state detector 4 sends an ON signal to the input controller 2 when the state of the switch 3 changes from OFF to ON, and changes the state of the switch 3 from ON to O.
An OFF signal is sent to the input control unit 2 when the state becomes the FF state.

For example, in the case of a speech recognition method using a Hidden Markov Model (hereinafter, referred to as HMM), the speech recognition unit 5 obtains the occurrence probability of all HMMs corresponding to the word to be recognized, and obtains the highest occurrence probability. A word corresponding to the HMM is set as a recognition result.

The HMM is a model of the statistical characteristics of speech obtained from a large amount of speech data.
Details of the speech recognition method using M are disclosed in "Speech Recognition by Stochastic Model" by Seichi Nakagawa.

The display unit 6 displays icons, recognition results, and the like.

The operation of input control in the above-structured speech recognition apparatus will be specifically described.

First, an example in which the user speaks while pressing the switch 3 to input a voice will be described with reference to the flowchart of FIG. 2 and FIG.

Step S11: The switch state detector 4 determines whether the user has pressed the switch 3. If it is determined that the switch 3 has been pressed, the input control unit 2 notifies the voice input unit 1 that the user wants to use the voice input unit 1 (step S12).

Step S13: By the input control unit 2,
It is determined based on the information from the voice input unit 1 whether it can be used. If the voice input unit 1 is being used by another application software, the voice input unit 1 cannot be used. Therefore, the process proceeds to step S22, and the other application software is currently using the voice input unit 1. An error message to the effect is output to the user, and the process ends.

If the voice input unit 1 can be used, the voice input unit 1 is thereafter used by another application software.
Cannot be used, and the input control unit 2 notifies the voice input unit 1 of the start of capturing of the voice signal (step S14).

Step S15: The voice input unit 1
An audio signal is captured.

Step S16: The input control unit 2
It is determined whether or not the elapsed time since the switch 3 was pressed (the elapsed time since ON) is longer than a preset time t0.

As a method of determining the set time t 0, for example, a voice input experiment is performed on the user under the condition “Please confirm that the voice input is possible and immediately start speaking.” As shown in FIG. 6A, there is a method of checking and setting a state in which a voice can be input and a timing of utterance. If the set time t0 from the time when the switch 3 is turned on until the voice input becomes possible is too long, the usability of the user deteriorates.
ms is desirable.

If it is determined in step S16 that the elapsed time from ON is shorter than the set time t0, the process returns to step S15 to continue capturing the audio signal. If the elapsed time from ON is longer than the set time t0, the process proceeds to step S17, and the input control unit 2 notifies the user that the voice can be input (see the timing chart of FIG. 5). .

As a method of notification, as shown in FIG. 4, for example, an icon as shown in FIG.
There is a method of displaying an icon like. The user speaks after confirming that the speech recognition device is ready for input.

Step S18: The switch state detector 4 determines whether or not the user has released the switch 3. When it is determined that the switch 3 is kept pressed,
Returning to step S15, the capture of the audio signal is continued.
If it is determined that the switch 3 has been released, the process proceeds to step S19, where the input control unit 2 determines that the elapsed time since the switch 3 was released (the elapsed time from OFF) is longer than the preset time t1. It is determined whether it is long.

As a method of determining the set time t 1, for example, the user is asked to “turn on the switch as soon as the utterance ends.
Please F. , An experiment of voice input is performed under the condition, and as shown in FIG. 6B, there is a method of checking and setting the timing of the end of the utterance and the switch OFF.

If it is determined in step S19 that the elapsed time from OFF is shorter than the set time t1, the process returns to step S15 to continue capturing the audio signal. If the elapsed time from OFF is longer than the set time t1, the process proceeds to step S20, and the input control unit 2 notifies the voice input unit 1 of the completion of the capture of the voice signal.

Step S21: By the input control unit 2,
The voice input 1 is notified that the use of the voice input unit 1 has been completed. Thereafter, other application software can freely use the voice input unit 1.

Next, an example of a case where it is determined whether the end of the capture of the audio signal is in the audio section or not will be described with reference to the flowchart of FIG. 3 and FIG.

Step S31: The switch state detector 4 determines whether the user has pressed the switch 3. If it is determined that the switch 3 has been pressed, the input control unit 2 notifies the voice input unit 1 that the user wants to use the voice input unit 1 (step S32).

Step S33: By the input control unit 2,
It is determined based on the information from the voice input unit 1 whether it can be used. If the voice input unit 1 is being used by another application software, the voice input unit 1 cannot be used. Therefore, the process proceeds to step S43, and the other application software is currently using the voice input unit 1. An error message to the effect is output to the user, and the process ends.

When the voice input unit 1 can be used, the voice input unit 1 is thereafter used by another application software.
Cannot be used, and the input control unit 2 notifies the voice input unit 1 of the start of capturing the voice signal (step S34).

Step S35: The voice input unit 1
An audio signal is captured.

Step S36: By the input control unit 2,
It is determined whether or not the elapsed time since the switch 3 was pressed (the elapsed time since ON) is longer than a preset time t0.

If the elapsed time from ON is shorter than the set time t0, the flow returns to step S35 to continue taking in the audio signal. If the elapsed time from ON is longer than the set time t0, the process proceeds to step S37, and the input control unit 2 notifies the user that the voice input is possible. The set time t0 is determined in the same manner as described above.

Step S38: By the input control unit 2,
The voice section is cut out.

The audio section is cut out, for example, after detecting a start point of the audio power calculated from the audio signal, at which the audio power has increased to a predetermined value or more, and detecting a temporary end point at which the audio power is equal to or less than the first threshold value. The temporary end point is sequentially moved to a position that exceeds the predetermined second threshold value and becomes less than the predetermined second threshold value, and the audio signal from the start point to the end point corresponds to the determination that the final temporary end point is the end point. Is performed in a process of cutting out.

Step S39: The switch state detector 4 determines whether or not the user has released the switch 3. When it is determined that the switch 3 is kept pressed,
Returning to step S35, the capture of the audio signal is continued.

If it is determined that the switch 3 has been released, the process proceeds to step S40, and after the input control unit 2 detects the start point, it is determined whether or not the end point of the voice section has been determined. If the end point has not been determined, step S
Returning to 35, the acquisition of the audio signal is continued. If the end point of the voice section has been determined, the process proceeds to step S41, and the input control unit 2 notifies the voice input unit 1 of the end of the capture of the voice signal.

Step S42: By the input control unit 2,
The voice input 1 is notified that the use of the voice input unit 1 has been completed.

By adding the processing described above to the preceding stage of the speech recognition processing, as shown in FIG. 5, even in the case where the speech at the end is missing in the conventional processing, the speech at the end is missing. Audio signals can be captured without the need.

That is, the user presses the switch 3 and
It is sufficient to remember that the user only speaks once after confirming the icon indicating the state in which the voice can be input, and it is not necessary to be so conscious of the timing of the utterance and the timing of releasing the switch. Furthermore, since the audio signal is captured shortly before the icon indicating the state where voice input is possible is displayed, even if the utterance is a little earlier than the display of the icon indicating the state where voice input is possible, the voice at the beginning of the word will be lost. Chipping can be reduced.

In this embodiment, the input control unit 2 notifies the use of the voice input unit 1 and notifies the end of the use of the voice input unit 1 that the voice input unit 1 is operated only when necessary and the voice signal is used. Therefore, the power consumption can be suppressed by stopping the circuit operation during other times.

Further, when the voice input unit 1 is used from a plurality of application software, FIGS. 7A and 7B
As shown in the example, each application software S1, S
2, S3 occupies the voice input unit 1 only when necessary, so that each application software S1, S2, S3 can use the voice input unit 1 in a time-division manner.

Here, the above-described processing is executed by a program, and the whole or a part of the program is recorded on a computer-readable recording medium such as a floppy (registered trademark) disk or a hard disk directly or via a communication line. It may be stored in a medium in advance, and installed and used as needed.

[0063]

As described above, according to the present invention,
In a method in which a user directly specifies a necessary voice section using a switch, it is possible to reduce the loss of the first part and the last part of the voice section to be extracted, and to provide the user with a switch press timing and a utterance timing. It is possible to reduce the troublesomeness of making the user aware of the situation as much as possible. Further, power consumption can be suppressed by operating the audio input device only when necessary and taking in an audio signal. Furthermore, one application software does not always occupy the voice input unit, and it is possible to use the voice input unit from other application software.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of an embodiment of a speech recognition device of the present invention.

FIG. 2 is a flowchart illustrating an operation of controlling capture of an audio signal based on elapsed time.

FIG. 3 is a flowchart showing an operation of controlling the capture of a voice signal by voice section detection.

4A is a diagram illustrating an example of an icon indicating a state where voice input is not possible, and FIG. 4B is a diagram illustrating an example of an icon indicating a state where voice input is possible;

FIG. 5 is a diagram showing a switch operation, an audio signal capturing section, and a timing of a captured audio signal.

6A is a diagram illustrating a method of determining a set time with respect to an elapsed time from a switch ON, and FIG. 6B is a diagram illustrating a method of determining a set time with respect to an elapsed time from a switch OFF.

FIG. 7A illustrates a case where a voice input unit is used from a plurality of application software, and FIG. 7B illustrates a process when a use request is issued to the voice input unit from a plurality of application software. FIG.

FIG. 8 is a diagram showing a schematic configuration of a conventional voice recognition device.

[Explanation of symbols]

 Reference Signs List 1 voice input unit 1A microphone 1B A / D converter 2 input control unit 3 switch 4 switch state detection unit 5 voice recognition unit 6 display unit

Claims (6)

[Claims] 1. A voice recognition device which captures a voice signal in response to a switch operation for designating capture of a voice signal and recognizes the captured voice signal, comprising: a voice input unit; and a switch state detecting an operation state of the switch. A detection unit, comprising an input control unit, the input control unit,
Notifies the voice input unit of the use of the voice input unit when the capture of the voice signal is necessary, and notifies the voice input unit of the end of the use of the voice input unit when the capture of the voice signal is no longer required. And switch is turned on from OFF
Immediately notifies the audio input unit of the start of audio signal capture when the status changes, and switches from ON to OFF.
A voice recognition device having a function of notifying the end of capturing of a voice signal with a delay of a predetermined time when a state is entered. 2. The input control section notifies the voice input section of use of the voice input section and start of capturing of a voice signal to the voice input section immediately when the switch is turned on from the OFF state. 2. The voice according to claim 1, further comprising a function of notifying the voice input unit of the end of the capture of the audio signal and the end of use of the audio input unit after a predetermined time when the state is turned off. Recognition device. 3. The input control unit, when the switch is turned from OFF to ON, notifies the audio input unit immediately of the start of the capture of the audio signal, and starts from when the switch is turned ON. 2. The speech recognition device according to claim 1, further comprising a function of notifying a user that a speech input is possible after a preset time has elapsed. 4. The input control section has a function of notifying the audio input section of the completion of capturing of an audio signal after a preset set time has elapsed since the switch was turned from ON to OFF. The voice recognition device according to claim 1, wherein 5. The input control section has a voice section detection function and captures a voice signal to the voice input section after the switch is turned from ON to OFF and the detection of the voice section is completed. 2. The speech recognition device according to claim 1, further comprising a function of notifying an end. 6. A computer-readable recording medium storing a processing program for retrieving an audio signal in response to a switch operation for designating the capture of an audio signal and recognizing the captured audio signal, wherein the recording of the audio signal is performed. Notifies the voice input unit of the use of the voice input unit when necessary, and notifies the voice input unit of the end of use of the voice input unit when the capture of the voice signal is no longer necessary, and turns off the switch. ON
Immediately notifies the audio input unit of the start of audio signal capture when the status changes, and switches from ON to OFF.
A computer-readable recording medium having recorded thereon a processing program for notifying the end of taking in of an audio signal after a predetermined time when the state is set.