JPH11126092A - Voice recognition device and on-vehicle voice recognition device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a voice recognition device capable of suitably dealing with the situation when a voice uttered by a user isn't recognized due to a bad ambient environment. SOLUTION: The user's voice is inputted from a microphone 16 to be recognized by a voice recognition part 18 through processing in an acoustic processing part 14. When the voice isn't recognized, a signal processing control part 12 specifies the cause of nonrecognition. When the voice is buried in the noise, it is judged that the cause of the nonrecognition exists in the ambient environment of the microphone 16. Then, a body ECU 30 controls an equipment deteriorating the ambient environment of the microphone 16 to improve the ambient enviforment. For instance, when a power window 32 is opened, this window 32 is closed. Since the ambient environment is improved, the voice uttered next can be recognized.

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 more particularly to a voice recognition device for recognizing a voice uttered by a user. The voice recognition device of the present invention is suitable for mounting on a vehicle.

[0002]

2. Description of the Related Art Conventionally, operation switches connected to various on-vehicle devices such as a power window and a headlight are installed in a vehicle cabin. A user operates the on-vehicle devices using the operation switches. I was However, it is desirable to make the device operation as easy as possible so that the user can concentrate on driving. In order to meet such a demand, it is considered effective to provide a voice recognition device. The device can be operated by uttering a voice command, thereby improving operability.

[0003] A voice recognition device is a device for recognizing voice uttered by a person, and it has been proposed to use a voice recognition device as an input device for various electronic devices. The voice uttered by the user is converted to digital data. "voice"
Is a sentence, word, character, symbol, number, and the like. In one example, cepstrum (cepust) is applied to digital audio data.
rum), and recognizes what the user said. For a speech recognition device, for example,
This is described in "Sound and Speech Engineering" (Sadahiro Furui, Modern Science Co., pp. 174-).

At present, it is difficult to correctly recognize voice with a probability of 100%, and therefore, it is well known that a talkback function is provided. The speech recognition device is provided with a speech synthesis device. When the speech is recognized, a synthesized speech indicating the recognition result is generated and output. The user knows the recognition result,
It is determined whether the recognition result is correct. When the recognition result is incorrect, the user utters a voice again and inputs the voice to the recognition device. With such a talkback function, a correct recognition result can be reliably obtained. A speech recognition device having a talkback function is disclosed in, for example,
No. 996 and JP-A-1-177192.

[0005]

In the speech recognition apparatus, there are cases where the speech cannot be recognized even though the user utters the speech (hereinafter referred to as non-recognition). The cause of the unrecognition may be in how the user speaks. For example, the unrecognition occurs when the user speaks too slowly. Further, the cause of the non-recognition may be in the environment around the voice input means such as a microphone. Typically, the user's voice is buried in ambient noise entering the microphone, resulting in unrecognition.

In Japanese Patent Application Laid-Open No. 1-177192, a handset is provided as voice input means which is detachable from the recognition device main body in order to prevent non-recognition due to ambient noise. When the user holds the receiver close to the mouth, voice recognition in a noisy environment is ensured. In other words, by providing a special structure called a receiver, voice recognition independent of the surrounding environment is realized.

However, in a vehicle, a user utilizes a voice recognition function while driving. In a configuration in which the user picks up the handset while driving, the advantage of the voice recognition device that can operate the in-vehicle device without using the hand is not sufficiently utilized. Another disadvantage is that a space for the handset must be provided in the vehicle interior. Therefore, providing a handset to prevent non-recognition of voice due to the surrounding environment is not a good idea for a vehicle. Further, it is desirable not only in a vehicle but also in other situations that a configuration like a receiver is not necessary from the viewpoint of operability and the space of the device.

SUMMARY OF THE INVENTION The present invention has been made in view of the above problems, and an object of the present invention is to provide a voice recognition apparatus capable of appropriately coping with a case where voice recognition is not performed due to a bad surrounding environment. .

[0009]

According to the present invention, there is provided a voice recognition apparatus comprising: a voice input unit for inputting a voice uttered by a user;
Voice recognition means for recognizing the input voice, cause identification means for specifying the cause of the non-recognition when the voice recognition means does not recognize the voice, and the surrounding environment of the voice input means And a surrounding environment improving means for performing an improving process for improving the surrounding environment.

Further, the vehicle voice recognition device of the present invention is a device mounted on a vehicle for recognizing a voice uttered by a user, comprising: voice input means for inputting a voice uttered by the user; Recognizing means for recognizing voice, cause specifying means for specifying a cause of non-recognition when voice non-recognition occurs in the voice recognizing means, and at least one in-vehicle device which affects the surrounding environment of the voice input means Device control means for controlling the on-vehicle device that is deteriorating the surrounding environment when the cause of the unrecognition is in the surrounding environment, To improve.

According to the present invention, when unrecognition occurs, the cause identification means checks whether or not the cause of the unrecognition is in the surrounding environment. The surrounding environment is typically related to noise. If the noise is loud, the user's voice is buried in the noise, and non-recognition easily occurs. If the cause of the non-recognition is the surrounding environment, the surrounding environment is improved by controlling the on-vehicle device. Thereby, the cause of the non-recognition is removed, and the occurrence of subsequent non-recognition can be reduced. The user can make the speech recognition device recognize the voice by the next utterance even when the voice uttered by the user is not recognized, and can use the recognition device more conveniently. Here, the operation and effects of the present invention have been described with particular reference to a vehicle voice recognition device, but the same applies to a case where the present invention is applied to a voice recognition device provided other than in a vehicle.

The above noise is noise for voice recognition. Therefore, the above-mentioned noises include those which are not usually called noises, such as audio output and air-conditioner blowing noise.

Further, some of the devices controlled by the present invention, such as an air conditioner, deteriorate the surrounding environment due to a sound emitted from the device itself. Some windows, such as windows, do not emit noise or the like, but deteriorate the surrounding environment. The control target device of the present invention includes both of the above. That is, the device control unit improves the surrounding environment by suppressing the noise generation itself or preventing the transmission of the noise by shielding or the like.

Preferably, the voice recognition device of the present invention outputs a guidance voice prompting a re-speak after the surrounding environment is improved. When the user re-voices according to the guidance, the re-voiced voice is input under the improved ambient environment.

[0015] Preferably, prior to the execution of the improvement of the surrounding environment, a guidance sound for notifying the execution of the improvement is output. This allows the user to know in advance the operation of the device for improving the environment, and the user can prepare for the operation of the device. The user can also prohibit the execution of undesired device operations.

[0016] Preferably, in the above-mentioned vehicle speech recognition device, after the re-utterance in accordance with the guidance voice is performed, the device control means controls the vehicle-mounted device operated to improve the surrounding environment. Return to the state before operation.

[0017]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention (hereinafter, referred to as embodiments) will be described below with reference to the drawings. FIG. 1 is a block diagram showing a speech recognition apparatus to which the present invention is applied. In the present embodiment, the speech recognition device 10
Is provided as an input device of the navigation ECU 100. The user can input a destination of navigation or the like by uttering voice.

The speech recognition apparatus 10 is provided with a signal processing control section 12 having a CPU. Signal processing control unit 1
An acoustic processing unit 14 is connected to 2, and a microphone 16 is connected to the acoustic processing unit 14. The voice uttered by the user is converted into an analog electric signal by the microphone 16 and input to the sound processing unit 14. In the acoustic processing unit 14, the audio data is converted into a digital signal. Further, the acoustic processing unit 14 performs data processing required for speech recognition. For example, window function processing or Fourier transform processing is performed, and the cepstrum of the audio data is obtained. The processed audio data is output to the signal processing control unit 12.

The signal processing controller 12 is connected to a trigger switch 26. The trigger switch 26 is operated by the user. When the trigger switch 26 is pressed, the signal processing control unit 12 captures data of a voice uttered thereafter from the microphone 16 via the sound processing unit 14.

As a modification, a so-called press-talk switch may be provided instead of the trigger switch 26. The voice uttered while the press-talk switch is kept pressed is transmitted from the microphone 16 to the signal processing controller 12.
It is taken in. Further, as another modification, the voice recognition device 10 may be a device of a constant recognition type. Even if there is no operation of the trigger switch or the press-talk switch, the presence or absence of utterance by the user is monitored and detected.

Further, a speech recognition unit 18 is connected to the signal processing control unit 12. The signal processing control unit 12 sends the voice data captured from the sound processing unit 14 to the voice recognition unit 18. The voice recognition unit 18 is a so-called DSP or microcomputer, and analyzes voice data uttered by the user. The voice recognition unit 18 has a recognition dictionary storage unit (not shown) that stores standard voice data corresponding to the voice to be recognized. The voice recognition unit 18 compares the user's voice data with the standard voice data, and recognizes the user's voice based on the comparison result. At this time, standard voice data closest to the voice data of the user is obtained by a method using a hidden Markov model or the like.

The recognition result of the voice recognition unit 18 is sent to the signal processing control unit 12. Thereby, the signal processing control unit 12
Can tell what the user has done. Note that the sound processing unit 14 and the speech recognition unit 18 only need to perform a speech recognition process according to a known principle. A speech recognition technique different from the above example may be applied, and any speech recognition technique can be applied to the present embodiment.

Further, a speech synthesis unit 20 is connected to the signal processing control unit 12. According to the instruction of the signal processing control unit 12, the speech synthesis unit 20 generates synthesized speech of various messages. The voice synthesis unit 20 has a message storage unit (not shown) storing message data, and voice is synthesized using the message data. The generated synthesized voice is output from the speaker 22 and transmitted to the user.

The signal processing control unit 12 performs data communication with the body ECU 30 using the in-vehicle LAN by using the communication control unit 24. The body ECU 30 controls various in-vehicle devices. In the drawing, a power window 32, a sunroof 34, an air conditioner 36, an audio device 38, and a door lock 40 are shown as a part of devices to be controlled. The body ECU 30 controls the operation of the devices by outputting control signals to these devices. Further, various signals for knowing the state of the vehicle are input to body ECU 30. FIG. 1 illustrates a speedometer 42, which sends vehicle speed data to the body ECU 30.

Next, the operation of the system shown in FIG. 1 will be described.
When the user presses the trigger switch 26, the signal processing control unit 12 instructs the speech synthesis unit 20 to synthesize a speech request message. As a result, for example, synthesized voice data “Please input a command” is generated.
This synthesized voice is output from the speaker 22, and a beep “beep” is output from the speaker 22.

The user who has heard the utterance request and the beep utters a voice. For example, a voice “destination” is uttered to input a navigation destination. The user's voice is input to the microphone 16, and the input signal is taken into the signal processing control unit 12 through data processing in the sound processing unit 14. The voice data is sent from the signal processing control unit 12 to the voice recognition unit 18. The voice recognition unit 18 performs a recognition process for comparing the voice data of the user with the standard voice data.
The recognition result is sent back to the signal processing control unit 12.

When the user's voice is recognized in this manner, the signal processing control unit 12 commands the voice synthesis unit 20 to generate a talkback message. The content of the recognized voice is repeated in the message for the talkback. For example, when the user utters “destination”, a synthesized speech “destination recognized” is generated. Talkback sound is output from the speaker 22 and transmitted to the user. If the recognition result is incorrect,
The user depresses the trigger switch 26 within a predetermined waiting time after the talkback. When the switch is depressed, a message requesting re-speaking and a beep sound are output, and the user's voice is input again. If the trigger switch 26 is not depressed, the signal processing control unit 12
The recognized voice (for example, “destination”)
Notify CU100. Here, the communication control unit 24 performs data communication using the in-vehicle LAN.

Next, with reference to the flowchart of FIG. 2, a description will be given of the processing in the case where the speech is not recognized. In the recognition processing by the voice recognition unit 18, when there is no standard voice data close to the input data, it is determined that voice non-recognition has occurred. For example, considering a case where a method of calculating a distance between input data and standard voice data using a determinant of voice data is adopted as a voice recognition method, when there is no standard voice data shorter than a predetermined distance, an error occurs. Recognition occurs. Considering the case where a stochastic method using a Hidden Markov model is adopted, non-recognition occurs when there is no standard voice data giving a likelihood larger than a predetermined value.

As shown in FIG. 2, when unrecognition occurs,
First, the signal processing control unit 12 determines whether or not the cause of non-recognition is the surrounding environment (S10). In the present embodiment, the surrounding environment is a state around the microphone 16 as a voice input unit. The determination method of S10 will be described later. If the cause of the unrecognition is not in the surrounding environment, it is determined whether the cause of the non-recognition is in the user's way of speaking (S
12). For example, when it is known that the input signal includes voice, but it is not clear what is being said, it is considered that the cause of unrecognition is in the way of speaking. If S12 is also NO, the cause is not known, so the signal processing control unit 12 instructs the speech synthesis unit 20 to generate a message for a re-speaking request. A message for requesting a resound is output from the speaker 22, and then a beep sound is output (S14).

In S12, if it is determined that the cause of the voice non-recognition is the user's way of speaking, as shown in FIG. A voice request message is generated and output (S16). In S16, a beep sound is output following the output of the reissue request message. FIG. 3 shows the contents of errors that occur due to the way of speaking by the user and the contents of talkback when each error occurs.

For example, No. The error content of 1 is "early start of voice (timing of voice is too early)". If a sound of a certain level or more has already been detected at the start time of the beep sound of the utterance request (start time of cutting out the sound from the input signal), the above error No. 1 is determined to have occurred. At this time, in the past, a talkback was simply made, "The utterance timing is too early." In the present embodiment, a talkback with guidance is performed, such as "Please speak a little later after the beep." Further, in the present embodiment, after the guidance-added talkback, the recognition-related processing is changed in accordance with the error content (S18). In S18, the error No. In response to 1, control is performed to make the beep sound larger than usual. The user who hears the above talkback hears the next loud beep. Then, the user utters a voice at a later timing than before according to the message.

For example, for example, Error content of 2
"The sound is long." If a predetermined level or more of voice is still detected at the time when a predetermined time has elapsed after the output of the beep sound (that is, at the time when a predetermined input period (voice cutout period) has ended), the above error No. 2 is determined to have occurred. At this time, in the past, a talkback was simply made that "the utterance was too long." In the present embodiment, a talkback such as "Please speak a little shorter after the beep" is performed. In addition, a talkback such as "Please speak within seconds after a beep." Further, in the present embodiment, as the change of the recognition-related processing (S18),
The audio cutout period is extended. As a result, a recognition process adapted to the user's way of speaking is performed, and it is possible to recognize a speech that could not be recognized last time.

For example, for example, Error content of 3
"The sound is low." If no sound above a certain level is detected within the input period after the beep sound, the above error No. 3 is determined to have occurred. At this time, in the past, a talkback "I could not recognize." In the present embodiment, a talkback such as "Please speak a little louder after the beep" is performed. Further, in the present embodiment, the threshold of the voice input is lowered as the change of the recognition-related processing (S18). As a result, the recognition process is adapted to the user's way of speaking, and the speech that could not be recognized last time can be recognized.

For example, for example, Error content of 4 is
"The sound is short." If the sound of a certain level or higher obtained during the input period after the beep is too short, the above error No. 4 is determined to have occurred. At this time, in the past, a talkback was simply made that "the utterance was too short." In the present embodiment, a talkback such as "Please speak a little more slowly after the beep" is performed. In addition, the same processing is performed for the remaining errors in FIG.

Next, returning to FIG. 2, a description will be given of the processing when it is determined in S10 that the cause of the voice non-recognition is in the environment around the microphone 16 (YES). As described above, the surrounding environment in the present embodiment is the microphone 1 as a voice input unit.
6. When the noise around the microphone 16 is loud, the surrounding environment is bad and unrecognition may occur. The noise is, for example, a wind noise coming from a window or a sound around the vehicle. Audio output and air-conditioner blowing noise are also noises for the voice recognition device. In particular, noise in a frequency range close to human voice (around 1 kHz) becomes a problem.

As described above, a beep ("beep") is output from the speaker 22 after the message of the generation request. When the ambient noise is above a certain level after this beep, for example, when a sound with a power above a certain level is detected but voice cannot be recognized, it is determined that unrecognition due to the surrounding environment has occurred. You. At this stage, it is not necessary to identify the in-vehicle devices that are deteriorating the surrounding environment. This is because it is difficult to identify the cause device from the noise data except for special noise such as audio output and road noise.

When S10 is YES, the signal processing controller 1
2 communicates with the body ECU 30 using the communication control unit 24,
Send requests for improvement of the surrounding environment. In response to the request for improvement, the body ECU 30 checks the state of each device to be controlled, thereby searching for a device causing deterioration of the surrounding environment from among the devices to be controlled (S22). And body EC
U30 determines the device operation necessary for the environmental improvement, and notifies the signal processing control unit 12 of the device operation. The signal processing control unit 12 instructs the voice synthesis unit 20 to generate a guidance voice for announcing a device operation. The generated guidance voice is output from the speaker 22 (S24). After the notice of the operation of the device, the body ECU 30 outputs the control signal to operate the cause device, thereby improving the environment around the microphone 16 (S26). End of environmental improvement is body E
The information is transmitted from the CU 30 to the signal processing control unit 12. Then, a re-speaking request message is generated by the voice synthesis unit 20, and this message and a beep sound are output (S2).
8).

For example, in S22, the body ECU 30
The state of the power window 32 is checked. When the power window 32 is open, the body ECU 30 determines that the power window 32 is a device that causes environmental degradation. Then, in S24, a guidance voice saying “Close window” is output, and subsequently, body ECU 30 closes power window 32. Similar processing is performed on the sunroof 34. If the sunroof 34 is open, the sunroof 34 is identified as one of the cause devices,
The guidance sound "Close sunroof." Is output.

For example, in S22, the body ECU 30
Checks the state of the air conditioner 36. If the air inlet switching mode is set to inside air circulation,
The body ECU 30 determines that the air conditioner 36 is a device that causes environmental degradation. This is because in the inside air circulation mode, a large blowing sound is generated as compared with the outside air introduction mode. Also, even in the outside air introduction mode,
When the blowout amount is large, it is determined that the air conditioner 36 is the cause device. Then, in S24, a guidance voice of "Stop the air conditioner" is output, and subsequently, the body ECU 30 stops the air conditioner 36 from blowing air.

For example, in S22, the body ECU 30
Checks the state of the audio device 38. When the audio device 38 is outputting music or the like and the volume is higher than a predetermined value, the body ECU 30
Is determined to be a device causing environmental degradation. And S
At 24, a guidance voice "Stop audio" is output, and then the body ECU 30 causes the audio device 38 to reduce the volume (mute).

In addition to the above examples, any device that affects the surrounding environment of the audio input of the microphone 16 is suitably controlled by applying the present invention. For example, road noise also degrades the surrounding environment. Therefore, it is preferable that a control device for controlling the suspension is provided as a part of the body ECU 30. This control device controls the suspension to reduce road noise. Further, the engine sound and the mission sound also penetrate through the wall of the cabin and reach the microphone 16, thereby deteriorating the surrounding environment of the microphone 16. Thus, for example, when the speed data input from the speedometer 42 is equal to or greater than a predetermined value, the engine is specified as a device causing environmental degradation. Then, a guidance voice such as "Please reduce the speed" is output. At this time, it is also preferable that the engine ECU as one of the in-vehicle device control devices lowers the vehicle speed using an engine brake or the like. Further, the control of the vehicle speed is also effective for reducing the road noise. Further, it goes without saying that the wheel brake may be controlled to reduce the vehicle speed.

Returning to FIG. 2, the user utters a voice again in accordance with the reissue request in S28. Due to the improved environment, the re-uttered voice is not buried in the ambient noise,
Thus, the speech is now recognized. Signal processing control unit 12
Notifies the navigation ECU 100 of the content of the recognition result. Then, the next voice input is performed as needed. In the case of the above example, after inputting the voice of "destination", input of the place name and the like is performed. After a series of voice input is completed, the signal processing control unit 12 or the navigation device 100 notifies the body ECU 30 of the completion of the voice input. The body ECU 30 returns the in-vehicle device operated in S26 to the original state. For example, if the window was closed in S26, this window is opened again. If the audio is muted in S26, the volume is returned to the original.

The preferred embodiment of the present invention has been described above. In the present embodiment, the signal processing control unit 12 functions as the cause specifying unit of the present invention, and specifies the cause of the unrecognized speech. Further, the body ECU 30 functions as the surrounding environment improving unit of the present invention, and the microphone 16 serving as the voice input unit.
Improve the surrounding environment. That is, the body ECU 30
Improves the surrounding environment by controlling devices that are deteriorating the surrounding environment of the microphone 16. According to the present embodiment, the cause of the non-recognition is removed by improving the surrounding environment, and the occurrence of the subsequent voice non-recognition can be reduced. Even if the user does not recognize his voice,
Next, the uttered voice can be recognized by the voice recognition device, and the recognition device can be used more conveniently.

Hereinafter, a modified example of this embodiment will be described. As described above, in the present embodiment, various on-vehicle devices that affect the environment around the microphone 16 are controlled. When there are a plurality of devices to be controlled, it is preferable to rank these devices. By operating only the device with the highest rank, the device operation can be minimized. Further, the number of devices to be controlled may be only one (for example, a window).

In the present embodiment, the speech recognition device
It is used as an input device of a navigation device.
The speech recognition device may be used for other purposes. For example, it is preferable to provide the voice recognition device of the present invention as an input device of the body ECU 30 in FIG. The user
The device to be controlled by the body ECU 30 can be operated using voice commands. For example, door lock 40
When the voice command for the operation of is not recognized,
Control for closing the power window 32 and the like are automatically performed.

The scope of the present invention is not limited to a voice recognition device for a vehicle, and the present invention can be applied to any voice recognition device. For example, the present invention is also applicable to an electronic device with a voice recognition device of a type that is fixedly installed at one place. When the cause of the non-recognition is in the surrounding environment, the environment is improved by controlling the peripheral devices. For example, by closing a shielding means such as a surrounding window or door, the environment is improved by noise shielding. Further, for example, by stopping other peripheral devices (such as an air conditioner), the environment is improved by suppressing noise generation itself.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating an overall configuration of an embodiment of the present invention.

FIG. 2 is a flowchart illustrating a process performed when non-recognition of voice occurs.

FIG. 3 is a diagram showing talkback contents when the cause of non-recognition is in the way of speaking by the user.

[Explanation of symbols]

Reference Signs List 10 voice recognition device, 12 signal processing control unit, 14 sound processing unit, 16 microphone, 18 voice recognition unit, 20 voice synthesis unit, 22 speaker, 30 body ECU, 32
Power window, 34 sunroof, 36 air conditioner, 38 audio equipment, 40 door lock, 42 speedometer, 100 navigation ECU.

Claims (7)

[Claims] 1. A voice input means for inputting a voice uttered by a user, a voice recognition means for recognizing the input voice, and a cause of the non-recognition when the voice recognition means does not recognize the voice. When the cause identification means to identify and the cause of unrecognition are in the environment around the voice input means,
And a surrounding environment improving means for performing an improving process for improving the surrounding environment. 2. The recognition device according to claim 1, wherein after the improvement of the surrounding environment, a guidance sound prompting a re-speak is output. 3. The recognition device according to claim 1, wherein prior to the execution of the improvement of the surrounding environment, a guidance sound for notifying the execution of the improvement is output. 4. A vehicular voice recognition device mounted on a vehicle and recognizing a voice uttered by a user, comprising: voice input means for inputting a voice uttered by the user; and voice recognition means for recognizing the input voice. And a cause specifying means for specifying a cause of the non-recognition when the voice recognition means does not recognize the voice, and a device control means for controlling at least one in-vehicle device which affects the surrounding environment of the voice input means. And wherein, when the cause of the unrecognition is in the surrounding environment, the device control unit controls the on-vehicle device that deteriorates the surrounding environment to improve the surrounding environment. Vehicle voice recognition device. 5. The voice recognition device for a vehicle according to claim 4, wherein after the improvement of the surrounding environment, a guidance voice prompting the user to re-speak is output. 6. The recognition device according to claim 5, wherein after the re-utterance according to the guidance voice is performed, the device control means operates the on-vehicle device operated to improve the surrounding environment. A vehicle voice recognition device for returning to a previous state. 7. The vehicle speech recognition device according to claim 4, wherein prior to execution of the improvement of the surrounding environment, a guidance sound for notifying the execution of the improvement is output.