JPH0844387A - Voice recognizing device - Google Patents

Abstract

PURPOSE:To change over the dictionary for specified speakers in accordance with a user dispensing with a complex operation in a voice recognizing device used as a voice input device in an atuomobile. CONSTITUTION:The dictionary for a voice recognition 43 has plural dictionaries for specified speakers and plural dictionaries for unspecified speakers. A data read-out part 25 reads out personal recognition data held in the personal recognition data holding part of ignition key. A recognition deciding part 45 performs voice recognition by using dictionaries for specified speakers and directories for unspecified speaker in the case the personal recognition data are present while selecting the dictionary corresponding to a specific speaker performing a voice input with the dictionary for the voice recognition 43 based on the personal recognition and by using dictionaries for unspecified speakers in the case the personal recognition data are not present.

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 used as a voice input device in an automobile.

[0002]

2. Description of the Related Art In recent years, voice recognition devices for recognizing voices spoken by humans as words have been put into practical use. It is conceivable that this voice recognition device is used as a voice input device when voice inputting a destination, instruction information, etc., in a car navigation system, for example. By the way, in the voice recognition, the specific speaker recognition that recognizes the voice of a specific person (specific speaker) is more preferable than the unspecified speaker recognition that recognizes the voices of multiple unspecified people (unspecified speaker). Has a high recognition rate. Therefore,
Even in a voice recognition device for recognizing an unspecified speaker, it is possible to form a dictionary for a specified speaker by learning a dictionary for storing characteristics of a voice, and particularly improve the recognition rate for the specified speaker. Conceivable.

[0003]

In such a voice recognition device, when it is desired to have a plurality of specific speakers, it is necessary to provide a dictionary for each specific speaker and switch the dictionary according to the user. . In this case, if the user operates a switch or the like to switch the dictionary, the operation becomes complicated. It is also conceivable that the voice recognition device recognizes the speaker to discriminate the user, but it is difficult to put it into practical use at the current technical level.

Therefore, an object of the present invention is to provide a voice recognition device used as a voice input device in an automobile,
An object of the present invention is to provide a voice recognition device having a plurality of dictionaries for a specific speaker and capable of switching the dictionaries for a specific speaker according to a user without requiring a complicated operation.

[0005]

According to a first aspect of the present invention, a personal identification data holding section is provided in an ignition key of an automobile and holds personal identification data for identifying a specific speaker who inputs a voice. Personal identification data reading means for reading the personal identification data stored in the personal identification data storage unit, and a plurality of specific speaker dictionaries provided for each of a plurality of specific speakers and storing the characteristics of the voice of each specific speaker. And voice input based on the personal identification data read by the personal identification data reading unit and the feature extraction unit that extracts the feature of the voice input by the voice input unit. A specific speaker dictionary corresponding to a specific speaker is selected, and the input voice is recognized using the specific speaker dictionary and the features extracted by the feature extraction means. And identification means, by provided to the speech recognition device used as a sound input device in a car, to achieve the above object. The speech recognition apparatus according to claim 2 is the speech recognition apparatus according to claim 1, further comprising a dictionary for unspecified speakers that stores the characteristics of the speech for unspecified speakers, and the recognition means is personal identification data. If there is, the dictionary for specific speakers and the dictionary for unspecified speakers selected based on this personal identification data are used, and if there is no personal identification data, the dictionary for unspecified speakers is used. . A voice recognition device according to a third aspect is the voice recognition device according to the first or second aspect, which is used as a voice input device of a navigation device.

[0006]

In the voice recognition device according to the present invention, the personal identification data reading means reads the personal identification data held in the personal identification data holding portion of the ignition key of the automobile. Further, the feature extraction means extracts the features of the voice input by the voice input means, and the recognition means selects a dictionary for a specific speaker corresponding to the specific speaker who inputs the voice based on the personal identification data. The input voice is recognized by using the special speaker dictionary and the features extracted by the feature extraction means. In the voice recognition device according to claim 2, when the recognition means has personal identification data, the personal identification is performed by using a specific speaker dictionary and an unspecified speaker dictionary selected based on the personal identification data. If there is no data, the input speech is recognized using the unspecified speaker dictionary. The voice recognition device according to claim 3 is used as a voice input device of a navigation device.

[0007]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT A preferred embodiment of the speech recognition apparatus of the present invention will be described in detail below with reference to FIGS. FIG. 1 is a block diagram showing the configuration of a navigation device using a voice recognition device according to an embodiment of the present invention as a voice input device. The navigation device includes a calculation unit 10, and the calculation unit 10 includes a display 11a that functions as a touch panel and the display 11a.
A display unit 11 including an operation switch 11b provided around the display unit, and a switch input management unit 12 that manages inputs from the touch panel of the display unit 11 and the switch 11b.
The current position measurement unit 13, the speed sensor 14, the map information storage unit 15, the voice recognition unit 16, and the voice output unit 17 are connected to each other.

The current position measuring unit 13 detects the current position where the vehicle is currently traveling or stopped by detecting coordinate data based on latitude and longitude. This current position measurement unit 1
GP which measures the position of the vehicle using the artificial satellite in 3
An S (Global Position System) receiver 21, a beacon receiving device 20 that receives position information from beacons placed on the road, an azimuth sensor 22, and a distance sensor 23 are connected, and the current position measuring unit 13 is connected to them. The current position of the vehicle is measured using the information of.

The direction sensor 22 is, for example, a geomagnetic sensor for detecting the direction of the vehicle by detecting the geomagnetism, a gas rate gyro or an optical fiber gyro for detecting the direction of the vehicle by detecting the rotational angular velocity of the vehicle and integrating the angular velocity. gyro,
A wheel sensor or the like in which left and right wheel sensors are arranged and the amount of displacement of the azimuth is calculated by detecting the turning of the vehicle based on the output pulse difference (difference in moving distance) is used. As the distance sensor 23, for example, various methods such as detecting and counting the number of rotations of the wheel or detecting acceleration and integrating twice are used. Although the GPS receiver 21 and the beacon receiving device 20 can measure the position independently,
In a place where reception by the PS receiver 21 or the beacon receiving device 20 is impossible, the current position is detected by dead reckoning using both the azimuth sensor 22 and the distance sensor 23.

A microphone 24 for inputting voice and a data reading unit 16 are connected to the voice recognizing unit 16.
The data reading unit 16 uses the ignition key 38 of the automobile.
Is provided in the insertion part of the. Ignition key 38
Is provided with a personal identification data holding unit 39 that holds personal identification data for identifying a specific speaker who inputs a voice. The personal identification data storage unit 39 stores the personal identification data that is different for each user depending on the form of the magnetization pattern, the bright and dark pattern such as a bar code, and the uneven pattern. , The personal identification data stored in the personal identification data storage unit 39,
The data is read, for example, magnetically, optically or mechanically depending on the form of data retention. The personal identification data holding section 39, the data reading section 25, the voice recognizing section 16 and the microphone 24 provided on the ignition key 38 constitute a voice recognizing apparatus of this embodiment.

The voice output section 17 includes a voice output IC 26 for outputting voice as an electric signal, and the voice output IC 2
An amplifier 27 that amplifies the output of 6 and a D / A converter 2 that digital-analog converts the output of the amplifier 27.
8 and. A speaker 29 is connected to the output terminal of the D / A converter 28.

The calculation unit 10 includes a map data reading unit 31 connected to the speed sensor 14 and the map information storage unit 15,
Map drawing unit 32, map management unit 33 managing map data reading unit 31 and map drawing unit 32, and map drawing unit 32
And a screen management unit 34 connected to the display unit 11, an input management unit 35 connected to the switch input management unit 12 and the voice recognition unit 16, and a voice output IC of the voice output unit 17.
Audio output management unit 36 connected to 26 and map management unit 3
3, a screen management unit 34, an input management unit 35, and an overall management unit 37 that manages the audio output management unit 36. The arithmetic unit 10 includes a CPU (central processing unit), a ROM (read only memory), a RAM (random access memory), etc., and the CPU executes a program stored in the ROM using the RAM as a working area. As a result, each of the above configurations is realized.

FIG. 2 is a block diagram showing the configuration of the voice recognition unit 16 in FIG. As shown in this figure, the voice recognition unit 16 includes a voice input unit 41 for converting a voice signal input from the microphone 24 into a digital signal, and a feature extraction for extracting a feature parameter from an output signal of the voice input unit 41. Unit 42, voice recognition dictionary 43 having an unspecified speaker dictionary and plural specified speaker dictionaries, and data reading unit 25.
Personal data storage for registering a standard pattern in the specific speaker dictionary by storing the characteristic parameters extracted by the characteristic extraction unit 42 as personal data in the specific speaker dictionary corresponding to the personal identification data read by And a processing unit 44. The characteristic parameter is a frequency characteristic of a voice signal in units of syllables, words, or phonemes. Further, the feature extraction unit 42 extracts this feature parameter by a multi-channel bandpass filter, a linear prediction analysis, or the like.

The voice recognition unit 16 further compares the standard pattern stored in the voice recognition dictionary 43 with the feature parameters extracted by the feature extraction unit 42 to recognize the input voice. 45 and this recognition determination unit 4
According to the content of the voice recognized by 5, the calculation unit 10
And a recognition result processing section 46 for generating an input signal to the input management section 35 of the calculation section 10. When there is the personal identification data read by the data reading unit 25, the recognition determination unit 45 uses the personal identification data from the voice recognition dictionary 43 to identify the specific speaker corresponding to the specific speaker who inputs the voice. A dictionary for personal speakers is selected, and the dictionary for specific speakers and the dictionary for unspecified speakers are used. If there is no personal identification data, the dictionary for unspecified speakers is used. When the registration mode is designated by the registration mode designating switch 40, the feature extraction unit 42 sets the extracted feature parameters to the personal data storage processing unit 4.
4 and outputs the extracted characteristic parameter to the recognition determination unit 45 in other cases. The personal data storage processing unit 44 operates when the registration mode is designated by the registration mode designation switch 40.

Here, an outline of the operation of the navigation device shown in FIG. 1 will be described. The driver inputs information such as the destination required by the navigation device by hand using the touch panel or the switch 11b of the display unit 11 or by voice using the microphone 24. Information input using the touch panel or the switch 11b is input to the input management unit 35 via the switch input management unit 12. Information input by voice using the microphone 24 is recognized by the voice recognition unit 16 and input to the input management unit 35.

When the arithmetic unit 10 prompts the driver to input information such as selection of a destination or a route, the screen management unit 34 displays the display 11a of the display unit 11 as necessary.
A message is displayed on the screen, or the voice output management unit 36 outputs voice guidance using the voice output unit 17 and the speaker 29.

The arithmetic unit 10 searches for a travel route to the destination based on information such as the destination input by the driver, and guides the route according to the travel route. At the time of this route guidance, the map management unit 33 inputs necessary map data from the map data reading unit 31 based on the measurement result of the current position measuring unit 13, and draws a map using the map drawing unit 32. Then, the screen management unit 34 displays the map on the display 11a. In addition, a traveling route, the current position of the vehicle, an arrow indicating the traveling direction, and the like are displayed on this screen.

Next, the operation of voice recognition will be described. Outline of main operation of voice recognition First, the voice recognition unit 16 reads from a voice recognition dictionary 43 in which various standard patterns for performing voice recognition are stored.
Identify the dictionary you need. That is, when the personal identification data is not read by the data reading unit 25, only the unspecified speaker dictionary is selected, as indicated by the thick line in FIG.
When the personal identification data is read, as indicated by the thick line in FIG. 3A, the corresponding specific speaker dictionary (FIG. 3A)
Then select the speaker A dictionary) and the unspecified speaker dictionary. Now, as shown in FIG. 3A, the dictionary for speaker A is
It is assumed that the standard pattern for the specific speaker A for the voices of “moving” and “left” has already been created from the characteristic parameters extracted by the characteristic extracting unit 42 based on the dictionary for the unspecified speaker. To do.

Here, a case where a voice "right", for example, as the voice of the specific speaker A is input from the microphone 24 to the voice input unit 41 of the voice recognition unit 16 will be described. Then, the feature extraction unit 42 extracts a feature parameter from the input "right" voice. Next, the recognition determination unit 45
Then, the extracted characteristic parameters are set to the standard patterns “movement” and “left” of the dictionary for the specific speaker A, and the standard patterns “movement”, “right”, and “stop” of the dictionary for the unspecified speaker. , “Left” and the similarity is calculated. Then, the voice having the highest degree of similarity, in this case, the standard pattern “right” of the unspecified speaker dictionary is recognized as the input voice. Since the voice recognition is performed using the standard pattern of the unspecified speaker dictionary instead of the specified speaker A dictionary, it is extracted by the feature extraction unit 42 as shown in FIG. 3B. The characteristic parameter of the voice "right" of the specific speaker A is registered in the dictionary for the specific speaker A as a standard pattern dedicated to the specific speaker A.

Details of Voice Recognition Operation Next, the details of the voice recognition operation described above will be described with reference to FIG.
This will be described using the flowchart of. Voice recognition unit 16
Determines whether or not there is personal identification data from the output of the data reading unit 25 in order to select a dictionary for use in voice recognition from the voice recognition dictionary 43 (step 20).
1). When there is no personal identification data (N), that is, when a new ignition key or master key having no personal identification data holding unit 39 is inserted, the unspecified speaker dictionary is selected from the voice recognition dictionary 43 (step 202). On the other hand, if there is personal identification data (step 201; Y), both the specific speaker dictionary and the unspecified speaker dictionary corresponding to the personal identification data are selected (step 203).

When a dictionary used for voice recognition is selected,
Next, it waits for a voice to be input from the microphone 24 (step 204), and when voice is input (step 204;
Y), the voice signal input from the microphone 24 is converted into a digital signal by the voice input unit 41, and the feature parameter is extracted from this signal by the feature extraction unit 42 (step 2).
05). Next, the recognition determination unit 45 compares the extracted feature parameter with each standard pattern stored in the voice recognition dictionary 43 selected in steps 202 and 203,
The input voice is recognized from the standard pattern with the highest similarity (step 206).

When the voice recognition (step 206) is completed, the voice recognition unit 16 displays a message confirming the recognized voice (word) on the display 11a or the speaker 2.
It outputs from 9 (step 207). The user recognizes the voice (word) recognized by the voice recognition unit 16 from this message.
If it is correct, press the confirmation switch. If it is not correct, press the re-input switch. The user may press the re-input switch only when it is determined that the message is incorrect. The voice recognition unit 16 determines whether or not it has been confirmed by the user (step 208),
If it is not confirmed (N), the process returns to step 204, and if it is confirmed (Y), the recognition result processing unit 46 performs recognition result processing (step 209). That is, the recognition result processing unit 46 generates an input signal to the calculation unit 10 according to the content of the voice recognized by the recognition determination unit 45, and supplies the input signal to the input management unit 35.

Next, the characteristic parameter extracted in step 205 is registered as a standard pattern in the specific speaker dictionary corresponding to the personal identification data (step 210), and the process returns to the main routine. On the other hand, if there is no personal identification data (step 202; N), the specific speaker dictionary has not been selected, so the process returns to the main routine without performing the standard pattern registration process in the specific speaker dictionary.

As described above, in the present embodiment, the voice recognition dictionary 43 is provided with a plurality of specific speaker dictionaries and non-specific speaker dictionaries, and the data reading section 25 holds the personal identification data of the ignition key 38. Since the personal identification data stored in the unit 39 is read and the recognition determination unit 45 selects the specific speaker dictionary corresponding to the specific speaker who inputs the voice based on the personal identification data, It is possible to switch the dictionary for a specific speaker according to a user without requiring a complicated operation. Further, according to the embodiment described above, it becomes possible to automatically register the standard pattern of each specific speaker in the specific speaker dictionary together with the voice recognition process.

Next, the voice recognition operation according to the second embodiment will be described. The configuration of the voice recognition device in the second embodiment is the same as that in the first embodiment. FIG. 5 is a flow chart showing the operation of registering a standard pattern in the specific speaker dictionary in the voice recognition device of this embodiment. This operation starts when the registration mode is designated by the registration mode designation switch 40 in FIG. In this operation, first, the data reading unit 25 reads the personal identification data held in the personal identification data holding unit 39 of the ignition key 38, and the personal data storage processing unit 44 acquires this personal identification data (step 30).
1). Next, under the control of the calculation unit 10, the word to be registered is output as a character from the display 11a or as a voice from the speaker 29 (step 30).
2).

Next, waiting for voice input from the microphone 24 (step 303), when voice is input (step 303; Y), the voice signal input from the microphone 24 is converted into a digital signal by the voice input unit 41. And the characteristic extraction unit 42 extracts characteristic parameters from this signal (step 304). Next, the personal data storage processing unit 44 stores the characteristic parameters extracted by the characteristic extraction unit 42 as personal data in the specific speaker dictionary corresponding to the personal identification data read by the data reading unit 25. , The standard pattern is registered in the specific speaker dictionary (step 305). It is also possible to input a background sound such as an engine sound when the voice is input from the microphone 24, and to register the characteristic parameter of the voice including the background sound as a standard pattern in the specific speaker dictionary. Next, it is judged whether or not there is the next word to be registered (step 306), and if there is (Y), the step 30 is executed.
Returning to 2, if there is no (N), the operation at the time of registration is ended. It should be noted that, in the specific speaker dictionary, only words that are particularly necessary among the words registered in the unspecified speaker dictionary may be registered.

FIG. 6 is a flow chart showing the operation at the time of voice input in the voice recognition apparatus of the second embodiment. In this operation, first, it is judged from the output of the data reading section 25 whether or not there is personal identification data (step 401).
When there is no personal identification data (N), that is, when a new ignition key or master key that does not have the personal identification data holding unit 39 is inserted, it is registered in the specific speaker dictionary or the unspecified speaker dictionary is used. A message requesting selection of whether or not to perform is output from the display 11a or the speaker 29 (step 402), and it is determined whether or not the user has selected registration (step 40).
3). If the user chooses to register,
A registration mode is designated by the registration mode designation switch 40. In this case (step 403; Y), the operation at the time of voice input is ended and the operation at the time of registration shown in FIG. 5 is performed.

When the user chooses to use the unspecified speaker dictionary, the user presses the switch for continuing the operation at the time of voice input. If there is personal identification data (step 4)
01; Y), and when the user selects to use the unspecified speaker dictionary when there is no personal identification data (step 403; N), the voice is next input from the microphone 24. When the voice is input (step 404; Y), the voice signal input from the microphone 24 is converted into a digital signal by the voice input unit 41, and the feature parameter is extracted from the feature signal by the feature extraction unit 42. Is extracted (step 405). Next, the recognition determination unit 45 compares the feature parameter extracted by the feature extraction unit 42 with the standard pattern registered in the voice recognition dictionary 43 to recognize the input voice (step 406).

On the other hand, when an ignition key such as a master key that does not have the personal identification data holding unit 39 is used, there is no personal identification data, and therefore the recognition determination unit 4
5 selects only the unspecified speaker dictionary from the voice recognition dictionary 43, as indicated by the thick line in FIG. 3 (c). Then, the recognition determination unit 45 recognizes the input voice by comparing the feature parameter extracted by the feature extraction unit 42 with the standard pattern stored in the unspecified speaker dictionary.

When the voice recognition (FIG. 6; step 406) is completed, the voice recognition unit 16 outputs a message confirming the recognized voice (word) from the display 11a or the speaker 29 (step 407). The user
From this message, if the voice (word) recognized by the voice recognition unit 16 is correct, the confirmation switch is pressed, and if not, the re-input switch is pressed. The voice recognition unit 16 determines whether or not it has been confirmed by the user (step 40).
8) If it is not confirmed (N), the process returns to step 404. If it is confirmed (Y), the recognition result processing unit 46 performs recognition result processing (step 409), and the operation at the time of voice input ends.

As described above, according to the second embodiment, the registration mode is designated without registering the standard pattern of each specific speaker together with the voice recognition process.
It becomes possible to independently register the standard pattern of a specific speaker.

In the embodiment described above, the standard pattern of the specific speaker is re-registered, and in step 302, the word to be registered is output as a character to the display 11a or is output as a voice from the speaker 29. However, prepare a standard pattern registration word table in advance,
The specific speaker may sequentially input voices to the microphone 24 according to this table. At this time, a reference number is attached to each standard pattern, and the specific speaker switches the reference number to the switch 1
You may make it the audio | voice recognition part 16 recognize the content of the audio | voice input after that from the microphone 24 by inputting from 1b. As a result, the specific speaker does not need to input voice for all words, and can select and register the necessary words.

FIG. 7 is a block diagram showing another example of the configuration of a navigation device using the voice recognition device according to one embodiment of the present invention as a voice input device. This navigation device does not use a display but guides a route by voice. Therefore, the display unit 11, the switch input management unit 12, the map drawing unit 32, and the screen management unit 34 in FIG. 1 are not provided. In addition, a switch 51 for instructing the operation of the navigation device, etc.
And a switch input management unit 52 that manages the input from the switch 51, and the switch input management unit 52 is connected to the input management unit 35. Other configurations of this navigation device are the same as those of the navigation device shown in FIG.

The present invention is not limited to the above-described embodiments. For example, the voice recognition device of the present invention is not limited to a navigation device, and may be used in winkers, power windows, rearview mirrors, antennas, car audio devices, etc. It can be used as a voice input device when other devices are operated by voice input.

[0035]

As described above, according to the present invention, a plurality of specific speaker dictionaries are provided, and the personal identification data reading means reads the personal identification data held in the personal identification data holding portion of the ignition key. Based on this personal identification data, the dictionary for the specific speaker corresponding to the specific speaker who inputs the voice is selected.Therefore, it is possible to select the specific speaker according to the user without complicated operations. You can switch the personal dictionary.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a navigation device using a voice recognition device according to an embodiment of the present invention as a voice input device.

FIG. 2 is a block diagram showing a configuration of a voice recognition unit in FIG.

FIG. 3 is an explanatory view showing an outline of a voice recognition dictionary in FIG.

FIG. 4 is a flowchart showing an operation at the time of voice input in this embodiment.

FIG. 5 is a flowchart showing an operation when registering a standard pattern in a specific speaker dictionary in the voice recognition device according to the second exemplary embodiment of the present invention.

FIG. 6 is a flowchart showing an operation at the time of voice input in the voice recognition device according to the second embodiment of the present invention.

FIG. 7 is a block diagram showing another configuration example of a navigation device using the voice recognition device according to the embodiment of the present invention as a voice input device.

[Explanation of symbols]

 10 arithmetic unit 16 voice recognition unit 24 microphone 25 data reading unit 38 ignition key 39 personal identification data holding unit 41 voice input unit 42 feature extraction unit 43 voice recognition dictionary 44 personal data storage processing unit 45 recognition determination unit 46 recognition result processing unit

Claims (3)

[Claims] 1. A voice recognition device used as a voice input device in a vehicle, the personal identification data holding unit being provided in an ignition key of the vehicle and holding personal identification data for identifying a specific speaker who inputs a voice. A personal identification data reading means for reading the personal identification data stored in the personal identification data storage unit; and a plurality of specific speakers provided for each of the plurality of specific speakers and storing the characteristics of the voice of each specific speaker. Dictionary, voice input means for inputting voice, feature extraction means for extracting features of voice input by this voice input means, and voice recognition based on personal identification data read by the personal identification data reading means. A dictionary for a specific speaker corresponding to a specific speaker to be input is selected and extracted by the dictionary for the specific speaker and the feature extracting means. Using features,
A voice recognition device, comprising: a recognition unit for recognizing an input voice. 2. A non-specific speaker dictionary for storing the characteristics of the voice for the non-specific speaker, wherein the recognizing means, if there is personal identification data, based on the personal identification data. 2. The voice recognition apparatus according to claim 1, wherein the selected speaker-specific dictionary and the speaker-unspecified speaker dictionary are used, and when there is no personal identification data, the speaker-specific speaker dictionary is used. 3. The voice recognition device according to claim 1, wherein the voice recognition device is used as a voice input device of a navigation device.