JPH11126088A - Device and method for recognizing voice, navigation device and navigation method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a device and method for recognizing voice rapidly recognizeing /processing inputted voice. SOLUTION: An acoustic analysis circuit 13 always analyzes voice signal from a handsfree microphone 11 regardless of switch setting of a selection switch 17 to supply a cepstrum coefficient etc., to the selection switch 17. The acoustic analysis circuit 16 always analyzes similarly the voice signal from the hand microphone 14 also regardless of the switch setting of the selection switch 17 to supply the cepstrum coefficient etc., to the selection switch 17. When the switching control of the selection switch 17 is performed, the cepstrum coefficient etc., is instantly supplied to a voice recognition circuit 18.

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, a voice recognition method, a navigation device, and a navigation method suitable for use in a vehicle-mounted navigation system.

[0002]

2. Description of the Related Art Today, navigation systems are provided which can be operated by voice. In such a navigation system, when a voice is input to a microphone provided near a driver's seat, a passenger seat, or a rear seat,
A predetermined operation can be performed.

[0003] However, there is a problem that the voice recognition rate is reduced due to the difference in characteristics of the microphones and the sound field. Therefore,
The applicant of the present application has proposed a "navigation system and a voice recognition device" which solves such a problem in Japanese Patent Application Laid-Open No. 6-274190.

Here, an outline of the voice recognition device of the navigation system will be described with reference to FIG. In the voice recognition device, the microphone 1 converts the input voice into a voice signal and supplies the voice signal to the terminal a of the switch 3.
Supplies an audio signal to the terminal b of the switch 3. The switch 3 is controlled to be switched to the terminal a or the terminal b by the system controller 10, and supplies the audio signal supplied to the terminal a or the terminal b to the analyzer 4. The analysis unit 4 obtains the power spectrum of the audio signal, normalizes the unique characteristics of the microphones 1 and 2 (microphone adaptation), and recognizes the power spectrum.
To supply. The recognizing unit 5 recognizes the voice signal by a predetermined process, and outputs the recognition result to the system controller 1.
Supply 0.

[0005] When the system controller 10 obtains a voice recognition result of "current position," for example, it obtains position information based on signals from the GPS antenna 6 and the GPS receiver 7, and determines the current location from the position information. A voice for explanation is output from the response unit 8 and the location is displayed on the display unit 9.

[0006]

The analysis unit 4
When the microphone adaptation process, particularly the noise subtraction process, is performed, there is a delay in the adaptation process and the calculation of the noise spectrum. Therefore, the analysis unit 4
However, immediately after the switching of the audio signal by the switch 3, the noise spectrum cannot be obtained immediately, and the response time is delayed, resulting in a problem that the overall processing speed is reduced. .

[0007] Such a delay in processing speed is a very serious problem for a user who needs to operate the car navigation system while driving a car, and urgent measures are required.

The present invention has been proposed in view of such circumstances, and a voice recognition device, a voice recognition method, a navigation device, and a navigation device capable of recognizing and processing input voice at higher speed. The aim is to provide a method.

[0009]

In order to solve the above-mentioned problems, a voice recognition apparatus according to the present invention comprises a plurality of voice signal conversion means for converting voices from a plurality of sound sources into voice signals, respectively, A plurality of sound analysis means for performing sound analysis of a sound signal output from the signal conversion means, a selection means for selecting and outputting one of the sound signals output from each sound analysis means, and an output from the selection means Voice recognition means for recognizing voice based on the voice signal obtained.

In the above speech recognition apparatus, the selection means selects any one of the audio signals output from each of the sound analysis means, and the speech recognition means recognizes the speech.

[0011] In the voice recognition method according to the present invention, voices from a plurality of sound sources are respectively converted into voice signals, acoustic analysis of each voice signal is performed, and any one of the voice signals subjected to the voice analysis is analyzed. Is selected and output, and voice is recognized based on the selected and output voice signal.

In the above-mentioned speech recognition method, one of the speech signals output from each acoustic analysis means is selected and the speech is recognized.

A navigation device according to the present invention comprises a plurality of audio signal conversion means for converting sounds from a plurality of sound sources into audio signals, respectively, and a plurality of audio signals for performing an acoustic analysis of the audio signals output from each audio signal conversion means. Sound analysis means, selection means for selecting and outputting any one of the sound signals output from each sound analysis means, sound recognition means for recognizing a sound based on the sound signal output from the selection means, Output means for outputting guidance information based on a voice recognition result of the voice recognition unit.

In the above navigation device, the selection means selects one of the audio signals output from each of the sound analysis means and recognizes the voice by the voice recognition means. The guidance information is output based on the recognition result.

In the navigation method according to the present invention, sounds from a plurality of sound sources are converted into sound signals, sound analysis of each sound signal is performed, and any one of the sound signals subjected to the sound analysis is analyzed. Is selected and output, a voice is recognized based on the selected and output voice signal, and guidance information is output based on the voice recognition result.

In the above navigation method, one of the audio signals output by analyzing the voice from each sound source is selected, and the voice is recognized, and the guidance information is output based on the voice recognition result. I do.

[0017]

Embodiments of the present invention will be described below with reference to the drawings. The first embodiment of the present invention can be applied to, for example, a navigation system 1 having a configuration shown in FIG.

The navigation system 1 is mounted on a car, and displays a current position of the car to indicate a route to a destination. The navigation system 1 can perform a predetermined operation by a voice of a driver or the like, and a voice recognition device 10 for recognizing the voice, and a navigation device 20 for displaying a current position or the like based on a recognition result of the voice.
It consists of.

The speech recognition apparatus 10 includes a hands-free microphone 11 provided in a driver's seat of a car, an analog / digital (A / D) converter 12 for digitizing a speech signal, and an acoustic analysis circuit 13 for analyzing a speech signal. A hand microphone 14 provided on the passenger side, an A / D converter 15, an acoustic analysis circuit 16, a selection switch 17 for switching and outputting an audio signal, and an audio recognition circuit for recognizing the content of the audio signal. 18.

The hands-free microphone 11 has a single directivity, converts a driver's voice into a voice signal, and supplies the voice signal to the A / D converter 12. In addition, a switch 11a is provided on the hands-free microphone 11. When the switch 11a is turned on, it supplies an on signal to a system controller 28 described later. In addition, the user,
Before emitting a sound to the hands-free microphone 11, it is necessary to turn on the switch 11a. Then, the A / D converter 12 digitizes the audio signal from the hands-free microphone 11 and supplies the audio signal to the acoustic analysis circuit 13.

As shown in FIG. 2, the acoustic analysis circuit 13
It includes a frequency analysis unit 31, an adaptation unit 32, a noise subtraction unit 33, and a feature extraction unit 34.

The frequency analysis unit 31 performs a frequency analysis of the audio signal using a general FFT as a frequency analysis method, and obtains a power spectrum of the audio signal.

The adaptation unit 32 stores microphone adaptation data specific to the hands-free microphone 11 in a memory, and executes a microphone adaptation process for normalizing characteristics specific to the microphone. The adaptation data is obtained so that the frequency characteristics of the hands-free microphone 11 become the characteristics of a flat standard microphone. That is, the microphone adaptation data differs depending on the type of the microphone, and when using a plurality of microphones, it is necessary to prepare each of them.

The noise subtraction unit 33 determines the spectrum of this noise section as a signal section other than the speech section as a noise section, and performs processing of subtracting this noise spectrum in the speech section.

If the power spectrum of the frequency analysis result is used as it is, the feature extraction unit 34 has many dimensions of the parameter, and the amount of calculation of the speech recognition circuit 18 increases. The feature extraction unit 34
For example, using a cepstrum, in particular, in speech recognition, a compression process is performed using a mel cepstrum in which an audible characteristic is weighted on a frequency axis, and a coded cepstrum coefficient and a mel cepstrum coefficient are output. To the terminal a.

On the other hand, the hand microphone 14 converts, for example, the voice of a person sitting in the passenger seat into an audio signal and supplies it to the A / D converter 15. The hand microphone 14 is also provided with a switch 14a. When the switch 14a is turned on, it supplies an on signal to the system controller 28. Then, the A / D converter 15 digitizes the audio signal, outputs audio data, and supplies the audio data to the acoustic analysis circuit 16. The acoustic analysis circuit 16 has the same configuration as the above-described acoustic analysis circuit 13, and supplies the analysis result of the audio data to the terminal b of the selection switch 17.

The selection switch 17 is controlled to be switched to a terminal a or a terminal b by a system controller 28. Specifically, the system controller 28 controls the switch 11
When the ON signal is supplied from a, the selection switch 17 is set to the terminal a, and when the ON signal is supplied from the switch 14a, the selection switch 17 is set to the terminal b. Then, the selection switch 17 supplies the coded cepstrum coefficient and the mel cepstrum coefficient supplied to the terminal a or the terminal b to the speech recognition circuit 18.

Here, the sound analysis circuit 13 always analyzes the audio signal from the hands-free microphone 11 and supplies the cepstrum coefficient and the like to the selection switch 17 irrespective of the setting of the selection switch 17. . Similarly, the acoustic analysis circuit 16 always analyzes the audio signal from the hand microphone 14 and supplies the cepstrum coefficient and the like to the selection switch 17 irrespective of the switching setting of the selection switch 17. Therefore, when the switching control of the selection switch 17 is performed, the cepstrum coefficient and the like are immediately supplied to the speech recognition circuit 18. That is, the selection switch 17 is. Even immediately after the terminal is switched, a cepstral coefficient or the like that is not delayed at all can be supplied to the speech recognition circuit 18.

The speech recognition circuit 18 calculates the hands-free microphone 1 based on the cepstrum coefficient and the mel cepstrum coefficient.
1 or the voice input to the hand microphone 14 is converted to HMM (Hi
ddenMarkov Model) Recognize based on the method. That is, the speech recognition circuit 18 uses the model (HMM) stored in the memory 18a in advance to
Alternatively, the voice input to the hand microphone 14 is recognized, and the recognition result is supplied to the system controller 28.

Further, even immediately after the selection switch 17 is switched, the speech recognition circuit 18 is supplied with a cepstrum coefficient or the like with no delay, so that the speech can be immediately recognized. Can be recognized faster than in the past.

The navigation device 20 has a G
PS antenna 21, GPS receiving circuit 22, CD-R
A CD-ROM drive 23 for reading map data and the like from an OM (Compact Disc-Read Only Memory), an autonomous sensor 24, a response unit 25 for outputting sound, a display unit 26 for displaying a current position and the like together with a map, A mobile phone connection unit 27 for connecting to a mobile phone not to be connected, and a system controller 28 for controlling the whole.

The GPS receiving circuit 22 receives a measurement signal from a GPS satellite via the GPS antenna 21 and analyzes the signal to detect a current position. Note that the data indicating the current position is data of the latitude and longitude at the place. The GPS receiving circuit 22 supplies the latitude and longitude data to the system controller 28.

When the position data is supplied from the GPS receiving circuit 22, the system controller 28
Control is performed such that data of a road map near the position indicated by the position data is read from the OM drive 23. Based on the road map data read from the CD-ROM drive 23, the system controller 28 displays a road on the display unit 26 or outputs a voice of the current location such as "currently XX" to the response unit 25. Let it.

The system controller 28 can input a place name by voice, switch the display mode of the display unit 26, and the like based on the recognition result of the voice recognition circuit 18. For example, when the driver turns on the switch 11a to generate a place name in the hands-free microphone 11, the system controller 28 recognizes the place name via the hands-free microphone 11, the acoustic analysis circuit 13, the voice recognition circuit 18, and the like. And the data of the place name is stored in the CD-ROM drive 23
, The location of the place name can be marked on the display unit 26.

On the other hand, the autonomous sensor 24 is supplied with a pulse signal corresponding to the vehicle speed supplied from an engine control computer (not shown) of the vehicle, and calculates an accurate running speed of the vehicle based on the pulse signal. The traveling direction is detected based on the output of a gyro sensor (not shown) inside the gyro sensor. Furthermore, the current position is measured by autonomous navigation from a position determined by the speed and the traveling direction.

The mobile phone connection unit 27 connects the mobile phone with the navigation system 1. Thus, the user can make a telephone call while talking with the hands-free microphone 11, for example.

Next, a second embodiment of the present invention will be described with reference to FIGS. Note that the same circuits as those in the first embodiment are denoted by the same reference numerals, and detailed description is omitted.

The navigation system 1A according to the second embodiment includes a voice recognition device 10A and a navigation device 20A.

The voice recognition device 10A includes a hands-free microphone 11 for converting a voice into a voice signal, an A / D converter 12 for digitizing the voice signal, an acoustic analysis circuit 13 for analyzing voice data, and a hand microphone 14. , A / D
A converter 15, a sound analysis circuit 16, a selection switch 17 for switching and outputting a speech signal, a continuous word speech recognition circuit 18b capable of recognizing a continuous word,
A memory 19 in which dictionary data for recognizing continuous words of the input voice is stored.

The continuous word speech recognition circuit 18b, for example,
Continuous speech words can be recognized by a two-stage DP algorithm, a level building (LB) algorithm, a one-pulse algorithm, or the like. On the other hand, the memory 19 stores a continuous numeric word dictionary 19a and a command dictionary 19b. As a result, the continuous word speech recognition circuit 18b reads data matching the analysis result from the memory 19 based on the analysis result of the acoustic analysis circuit 13, and recognizes only a numeral and a predetermined command. Specifically, the memory 19 has a continuous numeric word dictionary 19
As a, 0 (input voice: zero), 1 (one), 2
Data for recognizing continuous voices of (2), 3 (3), 4 (4), 5 (5), 6 (6), 7 (7), 8 (8), 9 (9) are stored. Have been. The command dictionary 19b includes insertion (sound), deletion (sound), all deletion (sound), search (key), telephone call (tsuwa), latitude (id), and longitude (key). Etc.)
And the like are stored.

It should be noted that, in order to allow a free input method of a telephone number, it is possible to input the voice of a numeral by a method without specifying the number of digits. However, if the number of digits is too large, the recognition rate decreases. Therefore, the maximum number of digits is specified within a range where there is no problem in practical use. For example, by recognizing the voice every six digits,
It is just a good separator for phone numbers. In addition, it is possible to specify a 3-digit or 4-digit number for the postal code and a 3-digit or 2-digit number for the latitude / longitude, and perform recognition with a highly accurate recognition rate.

On the other hand, the navigation device 20A
S antenna 21, GPS receiving circuit 22, CD-RO
An M drive 23, a response unit 25, a display unit 26, a mobile phone connection unit 27 connected to a mobile phone (not shown), a system controller 28, and a remote controller (not shown) (not shown). The system includes a remote control circuit 29 for transmitting a control signal to the system controller, a position information / response information generating circuit 30, and a memory 31 for storing place name data and the like.

The position information / response information generation circuit 30
The location information from the S receiving circuit 22 is supplied via the system controller 28, and based on the location information, the location name data 31a, postal code data 31b, telephone number data 31c, municipalities / larger-size data which match the location information from the memory 31. Read 31d. Then, the system controller 28 causes the response unit 25 to output the sound based on the read place name data or the like.

Here, the operation of inputting a telephone number by voice using the microphone 11 will be described. The system controller 28 displays the result of the numerical recognition on the display unit 26, and supports the user so that the user can visually recognize the result.

When the system controller 28 enters, for example, the telephone number search / call mode, the system controller 28 displays "phone number search / call" in the mode area 26a of the display unit 26 as shown in FIG.

When the user speaks "03" while turning on the switch 11a and turns off the switch 11b, the system controller 28 recognizes "03" through the voice recognition circuit 18 and Phone number area 26
“b” is displayed in “b”.

When the user speaks “3XXX” while turning on the switch 11a and turns off the switch 11b, the system controller 28 recognizes “3XXX” via the voice recognition circuit 18 and "3XXX" is displayed following "03" in the telephone number area 26b.

Similarly, the user speaks the last four digits "1234" while turning on the switch 11a, and
When 1b is turned off, the system controller 28 recognizes “1234” via the voice recognition circuit 18 and displays “1” next to “03” in the telephone number area 26b of the display unit 26.
234 "is displayed.

Note that, even when a wrong number is recognized by mistake, the user can correct the wrong number by voice. For example, the system controller 28
When the user speaks “Migi” to the hands-free microphone 11, the cursor on the display unit 26 moves one position to the right. Similarly, if you say "hidden," the curdle will move one position to the left. The cursor moves to the right when you say "Minagi" and the cursor moves to the left when you say "Minagiri". In addition, when "success" is said, only the number indicated by the cursor is deleted, and when "surge" is displayed, all the displayed numbers are deleted. The user can move the cursor in this way, delete the incorrectly recognized number, and then input the correct character again by voice.

When the user selects the search button 26c using a remote controller (not shown), the system controller 28 searches the memory 31 via the position information / response information generation circuit 30 for the telephone number data 31c. If there is latitude / longitude or name data corresponding to the telephone number, the display data and response synthetic voice data are read out, and a map of the place corresponding to the telephone number on the display unit 26 and the name of the place are displayed. Then, the response unit 25 outputs a synthesized voice of the place name of the place.

On the other hand, when the user selects the call button 26d using a remote controller (not shown), the system controller 28 supplies the telephone number to the portable telephone connection unit 27 and makes a call from the portable telephone. it can. At this time, the user can talk while using the hands-free microphone 11.

In the search mode based on the postal code, the system controller 28 searches the postal code data 31b from the memory 31 via the position information / response information generation circuit 30 based on the input voice. Then, the display data of the latitude / longitude information corresponding to the postal code and the information of the place name (city, town, village, capital letter name) indicated by the number and the synthetic data for response are obtained, and thereby the display unit 26 displays the information of the place. The map and the place name are displayed, and the response unit 25 outputs the sound of the place name.

Therefore, it is possible to recognize a telephone number or a postal code by continuous digit recognition, and it is possible to perform a telephone call with a mobile phone, a place search by a telephone number or a postal code, and the like.

In the above-described embodiment, the number of microphones is two.
Although the above description has been made with reference to the above example, the present invention is not limited to this, and it goes without saying that there may be three or more microphones.

[0055]

As described above in detail, according to the voice recognition apparatus and the voice recognition method according to the present invention, voices from a plurality of sound sources are respectively converted into voice signals, and acoustic analysis of each voice signal is performed. By performing any one of them and selecting any one of the sound signals subjected to the sound analysis and performing the sound analysis, even if a sound from any sound source is selected, the recognition of the sound due to the delay of the sound analysis is delayed. To prevent
Higher-speed processing can be performed.

According to the navigation device and the navigation method according to the present invention, sounds from a plurality of sound sources are converted into sound signals, respectively, and sound analysis of each sound signal is performed to select any one of them. By recognizing a voice and outputting guidance information based on the recognition result of the voice, no matter which sound source is selected, it is possible to prevent delay in output of guidance information due to delay in acoustic analysis. Thus, higher-speed processing can be performed.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a navigation system according to a first embodiment of the present invention.

FIG. 2 is a block diagram showing a configuration of an acoustic analysis circuit of the voice recognition device of the navigation system.

FIG. 3 is a block diagram showing a configuration of a navigation system according to a second embodiment of the present invention.

FIG. 4 is a diagram showing a display example of a display unit of the navigation system.

FIG. 5 is a block diagram showing a schematic configuration of a voice recognition device of a conventional navigation system.

[Explanation of symbols]

1 navigation system, 10 voice recognition device, 1
1 Hands-free microphone, 13, 16 sound analysis circuit, 1
4 hand microphone, 17 selection switch, 18 voice recognition circuit

──────────────────────────────────────────────────の Continued on the front page (51) Int.Cl. ⁶ Identification code FI G08G 1/0969 G08G 1/0969 (72) Inventor Eiji Yamamoto 6-35 Kita Shinagawa, Shinagawa-ku, Tokyo Inside Sony Corporation (72) Inventor Satoko Ikezawa 6-7-35 Kita-Shinagawa, Shinagawa-ku, Tokyo Inside Sony Corporation (72) Inventor Tomoaki Nitta 6-35-35 Kita-Shinagawa, Shinagawa-ku, Tokyo Inside Sony Corporation (72) Inventor Katsunori Miura 6-7-35 Kita Shinagawa, Shinagawa-ku, Tokyo Inside Sony Corporation

Claims (6)

[Claims] A plurality of sound signal converting means for converting sounds from a plurality of sound sources into sound signals; a plurality of sound analyzing means for performing sound analysis of a sound signal output from each sound signal converting means; A speech recognition apparatus comprising: a selection unit that selects and outputs any one of the speech signals output from each of the acoustic analysis units; and a speech recognition unit that recognizes speech based on the speech signal output from the selection unit. . 2. The sound signal converting means includes a switch, and the selecting means includes a switch that is turned on when any one of the sound signal converting means is turned on. 2. The speech recognition device according to claim 1, wherein the speech signal output by the device is selected and output. 3. Converting sounds from a plurality of sound sources into sound signals, performing sound analysis on each of the sound signals, and selecting and outputting any one of the sound signals on which sound analysis has been performed. A speech recognition method comprising recognizing speech based on a speech signal selected and output. 4. A plurality of sound signal conversion means for converting sounds from a plurality of sound sources into sound signals, a plurality of sound analysis means for performing sound analysis of a sound signal output from each sound signal conversion means, Selecting means for selecting and outputting any one of the sound signals output from the respective sound analyzing means; voice recognizing means for recognizing voice based on the sound signal output from the selecting means; Output means for outputting guidance information based on the result of voice recognition of the navigation device. 5. An audio signal conversion means comprising a switch, wherein each of the audio signal conversion means includes a switch, and wherein the selection means comprises a switch which is turned on when any one of the audio signal conversion means is turned on. 5. The navigation device according to claim 4, wherein the selected audio signal is output. 6. A sound signal from a plurality of sound sources is converted into a sound signal, sound analysis of each sound signal is performed, and one of the sound signals subjected to the sound analysis is selected and output. A navigation method comprising: recognizing a voice based on a voice signal selected and output; and outputting guidance information based on a voice recognition result.