JPH11338490A - Speech recognition unit for vehicle - Google Patents

Abstract

PROBLEM TO BE SOLVED: To eliminate the operation of a switch for starting speech recognition processing, and to enhance a speech recognition rate even in a noisy cabin. SOLUTION: A DSP 12 for an image inputs an image signal from a CCD camera 3 to operate an actual lip pattern P. The DSP 12 outputs a starring terminal signal P1 corresponding to a starting terminal t1 to a speech processor 15, based on the pattern P. A DSP 16 for a speech starts, from the starting terminal t1, speech recognition processing for a sound signal from a microphone 6 in response to the signal P1.

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 for a vehicle, and more particularly to cutting out a start end, a syllable end and an end in voice recognition processing.

[0002]

2. Description of the Related Art In recent years, with the spread of car navigation, a vehicle voice recognition device has been used to set a destination. Also, a voice recognition device for a vehicle is used for setting a telephone number in a hands-free telephone or the like. Therefore, further improvement in the voice recognition rate is required.

In general, in order to always start the voice recognition processing, it is very difficult to cut out the starting point of the voice of the driver from the noise in the cabin during traveling. Therefore, conventionally, an operation member has been provided in a vehicle voice recognition device to cut out the beginning of voice, and this operation member has been operated to start voice recognition processing.

[0004]

However, in these voice recognition devices, a switch operation for starting the voice recognition process is required, and even if the operation member is operated and the voice recognition process is started, the voice recognition process is immediately performed. Frequent occurrence of no voice input has made it difficult to detect the beginning of the voice. This has been a major cause of lowering the speech recognition rate.

[0005] Further, considerable noise is generated in the cabin during traveling. Therefore, it is required to further improve the speech recognition rate even in the noise. Also, in order to realize a true voice recognition device, it is necessary to eliminate the operation of operating the operation member to start the voice recognition process.

SUMMARY OF THE INVENTION An object of the present invention is to provide a vehicular speech recognition apparatus which eliminates a switch operation for starting speech recognition processing and can improve the speech recognition rate even in a noisy cabin. .

[0007]

According to a first aspect of the present invention, there is provided a microphone provided in a vehicle and a voice recognition means for recognizing a driver's voice detected by the microphone. An image pickup means provided in the vehicle, for picking up a face including the lips of the driver, and an image signal from the image pickup means, and the lip of the driver is obtained from the image signal. Calculates the lip pattern of each time, detects the start of the conversation of the pilot based on the lip pattern, inputs the voice of the pilot from the microphone to the voice recognition means, and performs voice recognition. The gist of the invention is that it comprises an image recognition means for outputting a start signal for starting a processing operation.

According to a second aspect of the present invention, in the vehicle speech recognition apparatus according to the first aspect, the image recognition means includes:
Based on the lip pattern, a syllable end from the start to the end of the conversation of the pilot is detected, and a syllable end signal is output.The voice recognition means responds to the start end signal based on the syllable end signal. After that, the gist of the present invention is that voice recognition processing is performed on the voice of the operator by categorizing the syllables based on the syllable end signal.

According to the first aspect of the present invention, the image recognition means calculates a lip pattern from the face imaged by the imaging means, detects the start of the conversation of the pilot based on the lip pattern, and outputs a start signal. Is output to the voice recognition means. The voice recognition means inputs the voice of the driver from the microphone in response to the start signal, and starts the voice recognition processing operation.

[0010] Therefore, it is possible to reliably and accurately identify the beginning of speech (start of conversation) without operating an operation member for identifying the beginning of speech (start of conversation) when performing speech recognition processing. Can improve speech recognition.

According to the second aspect of the present invention, the voice recognizing means divides the voice of the operator into syllables based on the syllable end signal from the image recognizing means, and performs voice recognition processing on the short voice of each syllable. Therefore, the voice recognition processing becomes easy. As a result, the voice recognition rate can be improved.

[0012]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of a vehicle voice recognition apparatus embodying the present invention will be described below with reference to FIGS.

As shown in FIG. 1, a CCD camera 3 as an image pickup means is provided on a pillar (A pillar) 2 in front of a driver's seat side in a vehicle cabin 1 so as to focus on a face of a pilot. ing.

A microphone 6 is provided on the pad 5 of the handle 4. The directivity of the microphone 6 is directed to the operator's mouth. Next, FIG. 2 shows an electrical configuration of the vehicle voice recognition device mounted on the vehicle.

The CCD camera 3 outputs an image signal obtained by differentiating the video image signal of the face including the lips of the pilot. This image signal is input to the image processing device 11. The image processing apparatus 11 includes an image digital signal processor (image DSP) 12 as image recognition means, a read-only and write-only memory (image RAM) 13 for temporarily storing data, and a read-only memory (image ROM) 1.
4. The image ROM 14 has an image D
A control program for performing high-speed digital operation by SP12 is stored. This control program
Transferred to the image RAM 13 in the image DSP 12,
Desired image recognition processing is performed in the image DSP 12.

The image DSP 12 receives an image signal from the CCD camera 3 and calculates an outline of the pilot's face from the image signal. Then, the image DSP 12 calculates the center position of the quadrilateral circumscribing the contour of the face and sets it as the nose position. Then, the image DSP 12 calculates a position obtained by adding a preset distance to the nose position and sets the calculated position as the lip center position.

The image DSP 12 fetches an image signal in a predetermined area around the lip center position and calculates a lip pattern (actual lip pattern) P. D for image
SP12 calculates the actual lip pattern P at each time at a preset time interval, and calculates the actual lip pattern P calculated at that time.
Is compared with the actual lip pattern P calculated immediately before. Then, for a preset time, both real lip patterns P
When there is no difference between the actual lip pattern P calculated at that time and the actual lip pattern P calculated immediately before,
The image DSP 12 outputs the start signal P1 corresponding to the start t1 of the voice recognition processing to the voice processing device 15.

When there is a difference between the actual lip pattern P calculated at that time and the actual lip pattern P calculated immediately before the image DSP 12, the image DSP 12 outputs the syllable end signal P2 corresponding to the syllable end t2 of the voice recognition processing. It is output to the audio processing device 15.

Further, when there is no difference between the calculated actual lip pattern P for a preset time, the image DSP 12 outputs a termination signal P3 corresponding to the termination t3 of the speech recognition processing to the speech processing unit 15. I have.

The microphone 6 outputs the voice of the operator to the voice processing device 15. The voice processing device 15 includes a voice DSP 16, a voice RAM 17, and a voice ROM 1.
8. The voice ROM 18 has a voice D
A control program for performing high-speed digital operation by SP16 is stored. This control program
The data is transferred to the audio RAM 17 in the audio DSP 16, and a desired audio recognition process is performed in the audio DSP 16.
The voice ROM 18 stores a plurality of standard voice command patterns including a plurality of voice command patterns set in advance.

The voice DSP 16 stores the plurality of standard voice command patterns in the voice RAM 17 in the voice DSP 16 when the vehicle is started by an ignition switch.
Read in.

As shown in FIG. 3, the voice DSP 16 is in a standby state in which the voice recognition processing is not started until the start signal P1 is input. The DSP 16 for audio is
When the start signal P1 from the SP 12 is input, it is determined that the conversation of the operator has started, that is, the voice DSP 16 determines that the voice start t1 for the voice recognition processing is input from the microphone 6. Then, the voice DSP 16 responds to the start signal P1 to start the voice recognition processing of the voice signal from the microphone 6 from the start point t1.

After the voice DSP 16 starts the voice recognition process based on the start signal P1, the voice DSP 16
When the syllable end signal P2 is input, one syllable is ended (syllable end t2) in the conversation of the pilot, and it is determined that the next syllable is to be transferred. Then, the voice DSP 16 divides the voice signal from the microphone 6 into sections corresponding to one syllable at each time in response to the syllable end signal P2, performs voice recognition processing on the separated voice signal, and the pilot has a conversation. Create real voice patterns. That is, the voice of the operator is classified into syllables based on the syllable end signal P2, and the voice recognition processing is performed on the short voice of each syllable. When the calculated actual voice pattern matches any one of a plurality of preset standard voice command patterns, the voice DSP 16 outputs a control signal corresponding to the matched standard voice command pattern.

Further, the audio DSP 16 is an image DSP 1
When the terminal signal P3 is input from the terminal 2, it is determined that the conversation of the operator has ended, that is, the terminal t of the voice for the voice recognition process
3 is determined to be input from the microphone 6. Then, after the voice DSP 16 receives the voice signal from the microphone 6 to the terminal t3 in response to the terminal signal P3, the voice DSP 16 does not receive the voice signal from the microphone 6 and enters a standby state in which the voice recognition process is not performed. And DS for audio
P16 waits for the input of the next new start signal P1.

In this embodiment, the image DSP 12 recognizes the conversation start, the syllable end, and the conversation end, respectively, and converts the start signal P1, the syllable end signal P2, and the end signal P3 into the audio DS.
The timing of the output to P16 is slightly earlier than the timing of the audio signal indicating the start, end of the syllable, and end of the conversation input from the microphone 6 to the voice DSP 16. Therefore, the voice DSP 16 can input the voice signal of the voice of the operator without delay and perform the voice recognition process.

If the timing at which the start signal P1, the syllable end signal P2, and the end signal P3 are output to the voice DSP 16 is much faster than the timing at which the voice signal is input, and there is a problem in voice recognition processing, the voice DSP 16 After inputting the start signal P1, the syllable end signal P2, and the end signal P3,
The operation for the voice recognition process may be started after a certain time.

According to the voice recognition device for a vehicle of the present embodiment, the following features can be obtained. (1) In the present embodiment, the image DSP 12
The lip pattern is calculated from the face imaged by the D camera 3, the start of the conversation of the pilot is detected based on the lip pattern, and the start signal P1 indicating the start of the conversation is output to the DSP 16 for voice. Then, the DSP 16 for audio outputs the start signal P
In response to 1, the voice of the operator is input from the microphone 6 and the voice recognition processing operation is started.

Therefore, it is possible to reliably and accurately specify the beginning of the voice (start of conversation) without operating an operation member for specifying the start of voice (start of conversation) when performing voice recognition processing. And the voice recognition rate can be improved.

(2) In this embodiment, the image DSP 12
The lip pattern is calculated from the face imaged by the CCD camera 3, and one syllable end t2 is detected in the conversation of the pilot based on the lip pattern, and a syllable end signal P2 indicating the syllable end t2 is sent to the voice DSP 16. Output. And
The voice DSP 16 responds to the syllable end signal P2 to classify the voice of the operator from the microphone 6 into syllables, and performs the voice recognition processing on the short voice of each syllable. Therefore, the voice recognition rate can be further improved.

(3) In this embodiment, the image DSP 12
A lip pattern is calculated from the face imaged by the CCD camera 3 to detect the end of the conversation of the pilot, and an end signal P3 indicating the end of the conversation is output to the DSP 16 for voice. Then, the voice DSP 16 performs voice recognition processing on the voice signal from the start t1 to the end t3 of the driver's voice in response to the end signal P3. Therefore, DS for audio
P16 performs only the voice recognition from the start to the end of the conversation of the operator, so that the voice recognition rate can be further improved.

(4) In this embodiment, the audio DSP 16
Performs a speech recognition process from the input of the start signal P1 to the input of the end signal P3, and otherwise enters a standby state in which the speech recognition process is not performed. Therefore, the voice DSP 16 does not perform voice recognition processing due to voice of a third party, noise from the outside of the vehicle, or the like. As a result, it is possible to prevent erroneous recognition of a voice caused by a voice of a third party or noise from outside the vehicle.

The embodiment of the present invention may be modified as follows. The image DSP 12 of the image processing apparatus 11 may be replaced with an image central processing unit (CPU).

The audio DSP 16 of the audio processor 15
May be changed to a central processing unit for voice (CPU). ○ An imaging tube camera may be used instead of the CCD camera 3.

A plurality of CCD cameras 3 may be used instead of one CCD camera 3. In this case, multiple CCs
One CCD camera 3 that accurately captures the lips of the driver is selected from the D cameras 3 and the selected CCD camera 3
The lip pattern is calculated using the image signal of.

A plurality of microphones may be used instead of one microphone 6. In this case, the sound from each microphone is input to the sound DSP 16 and synthesized.

Although the CCD camera 3 is provided on the A pillar 2, it may be provided on the inner mirror support 7. Instead of inputting an image signal obtained by differentiating the video image signal of the CCD camera 3 to the image processing device 11, the video image signal may be input to the image processing device 11. In this case, the video image signal input in the image processing device 11 is differentiated.

Even in the case of the configuration of each of the above-mentioned different examples, substantially the same features as those of the above embodiment can be obtained. Next, technical ideas other than the inventions described in the claims that can be grasped from the embodiment and other examples will be described below together with their effects.

(1) In the voice recognition apparatus for a vehicle according to claim 1, the image recognition means (12) detects the end of the conversation of the pilot based on the lip pattern, and generates an end signal (P3). Output, voice recognition means (16)
Receives the end signal (P3) and receives the start signal (P3).
A voice recognition device for a vehicle, which performs voice recognition processing on a voice of a driver during a period from responding to 1) to an end signal (P3).

Therefore, according to the invention described in (1), the speech recognition rate can be improved. (2) The vehicle voice recognition device according to claim 1,
The voice recognition means for a vehicle, wherein the voice recognition means (16) is in a standby state in which voice recognition processing is not performed except for a period from a start signal (P1) to an end signal P3 of the image recognition means (12). apparatus.

Therefore, according to the invention described in (2), erroneous recognition of voice can be prevented. (3) A voice recognition method for a vehicular voice recognition device, comprising: a microphone (6) provided in a vehicle; and voice recognition means (16) for recognizing a driver's voice detected by the microphone (6). An image of the face including the lips of the pilot is imaged, the image area of the lips of the pilot is cut out from the image signal, the lip pattern at each time is calculated, and the start of the conversation of the pilot is detected based on the lip pattern. A voice recognition method for a vehicular voice recognition device for inputting a voice of a driver from a microphone (6) to the voice recognition means (16) based on a start time thereof and performing a voice recognition processing operation.

[0041]

As described above in detail, according to the first aspect of the present invention, the voice recognition means can reliably output the voice from the microphone without operating the operation member for specifying the start of the voice. Since the voice recognition processing can be performed by capturing the voice, the voice recognition rate can be improved.

According to the second aspect of the present invention, the voice recognition processing is facilitated. As a result, the voice recognition rate can be improved.

[Brief description of the drawings]

FIG. 1 is a schematic diagram of a passenger compartment illustrating an arrangement of a CCD camera and a microphone according to an embodiment.

FIG. 2 is an electrical configuration diagram of the vehicle voice recognition device according to the embodiment.

FIG. 3 is a timing chart of an audio waveform in the embodiment.

[Explanation of symbols]

P1: start signal, P2: syllable end signal, t2: syllable end, 3
... CCD camera as imaging means, 6 ... Microphone, 12 ... Image DSP as image recognition means, 16 ...
DSP for voice as voice recognition means.

Continuation of the front page (51) Int.Cl. ⁶ Identification code FI G06T 1/00 G06F 15/62 380

Claims (2)

[Claims] 1. A microphone (6) provided in a vehicle
And a voice recognition means (16) for recognizing the driver's voice detected by the microphone (6). An image pickup device provided in the vehicle for capturing a face including the pilot's lips. Means (3), an image signal from the image pickup means (3) is input, an image area of the lips of the pilot is cut out from the image signal, and a lip pattern at each time is calculated, based on the lip pattern. Detecting the start of the conversation of the pilot, and recognizing said voice recognition means (16)
, A pilot's voice is input from the microphone (6), and a start signal (P) for starting a voice recognition processing operation is input.
1. A vehicle voice recognition device comprising: an image recognition means (12) for outputting 1). 2. The vehicle speech recognition device according to claim 1, wherein the image recognition means detects a syllable end (t2) from the start to the end of the conversation of the pilot based on the lip pattern. And outputs a syllable end signal (P2). The voice recognition means (16) outputs the syllable end signal (P2).
A voice recognition device for a vehicle, which responds to the start signal (P1) based on the syllable end signal (P2) and classifies the syllable based on the syllable end signal (P2) to perform voice recognition processing.