JP2003140682A - Voice recognition device and voice dictionary generation method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To automatically generate a voice dictionary of words that a user personally requires and to generate a voice dictionary which is reliable. SOLUTION: A voice dictionary database 52 in which words and voice patterns are registered corresponding to each other is prepared, and a voice dictionary generation part of a voice recognition engine 51 sends a voice pattern inputted from a microphone to an external voice recognizing device 70; and the correspondence between a word whose voice is recognized by the external voice recognizing device and the voice pattern of the word is registered in the voice dictionary database 52 and a voice recognition part of the voice recognition engine 51 recognizes the word of the inputted voice by using a voice dictionary registered in the voice dictionary database.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a voice recognition device and a voice dictionary creating method, and more particularly, it registers voice patterns of a plurality of words in a voice dictionary database and registers them as input voice patterns. The present invention relates to a voice recognition device for recognizing a voice input word based on the most similar registered voice pattern by comparing voice patterns and a voice dictionary creating method thereof.

[0002]

2. Description of the Related Art A navigation system searches for a route to a destination when a destination is input and guides a driver along the route. To enter the destination, (1) key in the address of the destination (address input), (2) if the destination is an intersection, specify two streets and enter the intersection (intersection Input), (3) is the destination
How to key in the name of POI (Point of Interest) (PO
(Enter I), (4) Enter the category and city name to display the POI candidate list, and select the PO that is the destination from the candidate list.
How to select I, (5) Directly input the destination on the map with the cursor, (6) Add a predetermined name to the address, register it in the address book, and display the address book And enter the destination address, (7) PO
There is a method of inputting the telephone number of I, etc.

9 and 10 are explanatory views of a destination specifying method by inputting an address. US address notation is state /
It has a format of city name / street name / address. Therefore, the destination can be specified by inputting the state / city name / street name / address. Therefore, in order to specify the destination by inputting an address in the navigation device, the user operates the menu key of the remote controller to display the main menu on the screen, and selects the menu item "Dest" from the main menu. By selecting "Dest", the navigation device displays the "Find Destination by" screen for specifying the destination input method as shown in Fig. 9 (a). Select the menu item "Address" and select When the enter key is pressed, the navigation control section displays the menu item "Change Stat" as shown in Fig. 9 (b).
e "," Street Name "," City Name "are displayed. To enter the destination address starting with the city name, select the menu item" City Name "and enter the destination address starting with the street name. To do this, go to the menu item "Street
Enter "Name". Also, since the last selected state is displayed as the default state, if you want to change the state name, select "Change State" (Fig. 9 (c)) and scroll to the desired state. Select the state (Fig. 9 (d)).

When the menu item "City Name" is selected and the enter key is operated after the state is determined, the navigation control section displays an alphabet / numeric keyboard on the screen for inputting the city name as shown in FIG. 9 (e). indicate. In this state, if the user enters the first few spells of the city name and enters the list key List on the keyboard,
As shown in FIG. 9 (f) and FIG. 10 (g), a list of city names having the spell in California is displayed in a scrollable manner. The figure shows the case where the spelling "T" is entered,
The city name "TORRANCE" with "T" at the beginning is displayed. Then, use the joystick keys to scroll through the city name list, select the desired city name (TORRANCE), and operate the enter key.

When the city name specifying operation is completed as described above, the navigation device displays the street name input screen as shown in FIG. 10 (h). In this state, the user enters the first few spells of the street and enters the list key on the keyboard.
Press List (Fig. 10 (i)). As a result, FIG. 10 (j)
As shown in, a list of street names existing in the TORRANCE having the spell is scrollably displayed.
The figure shows the case where the spelling "CAR" is entered, with "CAR" at the beginning
The street name "CARSON" with is listed. Then, use the joystick keys to scroll the street name list, select the desired street name (CARSON), and operate the enter key. When the street name identification work is completed as described above, the navigation control unit
The address input screen is displayed as shown in FIG. In this state, if you enter the address and operate the enter key,
The destination address is displayed as shown in Fig. 10 (m). If the address is correct as the destination, "OK to Proce
ed "is selected and confirmed as the destination. When the address of the destination is input, the navigation control unit obtains the latitude and longitude of the destination address from the address database and performs route search using the position information.

FIG. 11 is an explanatory diagram of a destination specifying method by inputting a POI type (category). In the navigation device, in order to specify the destination by the POI type input, the user operates the menu key of the remote controller to display the main menu on the screen, and selects the menu item "Dest" from the main menu. By selecting "Dest", the navigation controller displays the "Find Destination by" screen for specifying the destination input method, as shown in Fig. 11 (a). Select the menu item "Point of Interest". Then press the enter key on the remote controller and the navigation device will display the menu item "Plac" as shown in Fig. 11 (b).
e Name "," Place Type sort by distance "," Place Type
From within "a city" is displayed, select the menu item "Place Type sort by distance" and press the enter key, which causes the navigation controller to scroll the main category list as shown in Fig. 11 (c). You can change the contents of the list by displaying it and scrolling as shown in Fig. 11 (d) .There are some main categories with and without sub categories, RESTAURANT, SHOPPING, and GAS. S
TATION and the like have subcategories. When these are selected, for example, when a restaurant is selected, a subcategory list is scrollably displayed as shown in FIG. 11 (e).

When a predetermined POI type (category) is specified, the navigation controller displays the facility list of the category in a scrollable manner in the order of distance, as shown in FIG. 11 (f). Use the joystick keys on the remote control to scroll through the POI name list and select the desired POI (DEL AMO FASHION CE
If you select (NTER) and operate the enter key, the navigation device displays the destination address as shown in FIG. 11 (g). If the address is correct as the destination, "O
Select "K to Proceed" to confirm the destination. When the destination address is entered, the navigation device will
The latitude and longitude of I are obtained from the POI database, and a route search is performed using the position information.

By the way, as a recent tendency, it is possible to specify a destination by voice input instead of key input. FIG. 12 is an explanatory diagram in the case of inputting a destination address by voice. When the user presses the talk switch on the remote control (step 101), the system makes a "beep" sound. It is possible to input voice by the generation of this "beep" sound.
(The same applies hereinafter), the user inputs by voice "menu" (step 102). The navigation control unit recognizes the input voice and displays the "Find Destination by" screen on the screen to identify the destination input method as shown in Fig. 9 (a).
"ould you like to set your destinatio?" is output by voice, whereby the user inputs "Address" by voice (step 103).

The navigation controller recognizes the input voice and displays the menu item "Cha" on the screen as shown in FIG. 9 (b).
"F with" nge State "," Street Name ", and" City Name "
IndAddress by "Display the screen and select" Select City or Street "
"find your address in California." is output. The last selected state is displayed as the default state. If you want to change the state name, click "Change.
Input "State" by voice and click "Select a S" as shown in Fig. 9 (d).
The "tate" screen is displayed, and then the state name is input by voice. Note that the cancel can be returned to the state shown in FIG.

After the state is determined, the user voice-inputs "City" to input the destination address, starting with the city name.
(Step 104). The navigation control unit displays the "Input City Name" screen on the screen by displaying the alphabet / numeric keyboard for recognizing the input voice and inputting the city name on the screen as shown in FIG.
ich City in California? Is output as a voice (step 1
05). When the user enters the city name "TORRANCE" (step
106), the navigation controller displays the "Input Street Name" screen on the screen, which has an alphabet / numeric keyboard to recognize the input voice and input the street name on the screen as shown in FIG. 9 (h). Which Street in Torrance? "Is output as a voice (step 107).

When the user inputs the street name "Carson" (step 108), the navigation device recognizes the input voice and inputs "Address Address" on the screen as shown in FIG. 9 (k). Number "
Display the screen on the screen, and then "Which Street number on
Carson? "Voice out (step 109). User addresses"
When "2525" is input by voice (step 110), the navigation controller displays the "Confirm Destination" screen having the destination address as shown in FIG.
you like to calculate a route to 3525 Carson Torra
nce ST, CA? "is output by voice. If the address is correct as the destination, the user inputs" OK "by voice to determine the destination.
ing route to 3525 Carson ST Torrance, CA "is output by voice, the latitude and longitude of the destination address are obtained from the address database, and the route search is performed using the position information.

FIG. 13 is an explanatory diagram of the voice recognition device. The voice recognition device 1 is connected to, for example, the navigation device 2, and recognizes a word that the user has input by voice and inputs it to the navigation device 2. Speech recognition dictionary 1
a stores a word (Text) and a phonetic symbol number (Annotation),
The synthetic speech table 1b stores the correspondence between phonetic symbol numbers (Annotation) and phonetic symbols. Voice recognition engine (Voice Re
cognition Engine) 1c has the highest similarity (highest score) to the voice pattern of the speaker input from the microphone 1d by referring to the voice recognition dictionary 1a and the synthetic voice table 1b.
The word is determined and output to the navigation device 2. That is, the words in the voice recognition dictionary 1a that have been registered in advance are subjected to pattern matching with the input voice pattern, and the one with the highest score (probability) is taken as the recognition result.
As the recognition result, Text and Annotation corresponding to it
Output (ID).

[0013]

In order to perform voice recognition, it is necessary to register the word to be recognized in the voice recognition dictionary in advance. Address in navigation device
To enter the destination in (Address), you need to enter the city, street, etc. by voice, and in POI (Facility), enter the POI name.
(Facility name) needs to be input by voice. However, the names of these cities, streets, and facilities are very large, and there are hundreds of thousands. For this reason, the voice recognition dictionary becomes too large to correctly recognize the destination, and the hardware scale becomes large.
There are problems that the device becomes large and the cost becomes high.

Further, a commercially available voice recognition engine has a capability of a voice recognition target word having a maximum of 4000 utterances in 2 seconds, and cannot register all city names, street names and facility names. Therefore, the applicant of the present application, as Japanese Patent Application No. 2001-192737, internally inputs a region name or facility name that the user really needs in advance, and internally creates a voice pattern of the keyed region name or facility name. Then, a voice recognition device is proposed in which the voice pattern is registered in a voice dictionary in association with a region name or a facility name, and voice recognition is performed using the voice dictionary. However, this proposed voice recognition device has a problem that it is difficult to create a voice dictionary because each voice recognition target word must be registered by key input, and moreover, the voice pattern is internally generated from the key input data. The problem is that reliability is low because it is synthesized.

From the above, it is an object of the present invention to enable a user to automatically create a voice dictionary of words that he or she personally needs, and also to create a highly reliable voice dictionary for voice recognition. It is to be.

[0016]

According to the present invention, there is provided a voice dictionary database for registering a correspondence between a word and a voice pattern, and a voice pattern of an input word is sent to an external voice recognition device. A voice dictionary creating unit for registering a correspondence between a word recognized by an external voice recognition device and a voice pattern of the word in a voice dictionary database, and inputting using a voice dictionary registered in the voice dictionary database. This is achieved by a voice recognition device provided with a voice recognition unit that recognizes a word of a generated voice. That is, a highly reliable voice dictionary can be created.

Further, a dictionary unit for registering a subordinate concept word and its voice pattern as a set is provided for each superordinate concept, and when the subordinate concept word is input by voice, the word obtained by the external voice recognition device and its voice are obtained. The pattern is registered in the dictionary unit according to the superordinate concept. In this way, only the words registered in a predetermined dictionary unit need to be searched, and the number of words to be pattern-matched can be reduced. Therefore, high-speed voice recognition can be performed.

When a voice is input, the word recognition processing of the input voice is performed using the created voice dictionary. If the word can be recognized, the external voice recognition device is not inquired. Only when the word cannot be recognized, the external voice recognition device is used. The correspondence between the word recognized by the external voice recognition device and the voice pattern of the word is registered in the voice dictionary database. By doing so, it is possible to automatically create a voice dictionary in which voice patterns of words required by the user are registered while performing voice recognition.

[0019]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 is an overall configuration diagram of a system for realizing a voice dictionary creating method of the present invention. 10 is a navigation device, 50 is a voice recognition device, 70 is an external server for voice recognition, and 80 is a large-capacity dictionary database in which voice patterns of all words are registered. The voice recognition device 50 is a voice recognition engine unit 51 that performs voice recognition processing and voice dictionary creation processing.
It has a personal voice dictionary database 52 in which a personal voice dictionary for a user is registered, and a wireless communication section 53 for communicating with an external server 70 via a wireless communication network / internet. The voice recognition device 50 and the navigation device 10 are connected so that they can communicate with each other, and the navigation operation can be performed by voice.

When a predetermined word is voice-input on a predetermined navigation screen, the voice recognition engine 51 sends the voice waveform data (PCM) of the word to the external server 70 to cause the external server 70 to perform voice recognition processing. The speech recognition result (recognized word and phonetic symbol string) is acquired from an external server. In addition, the voice recognition engine 51 uses the acquired words as a navigation device.
The word and the voice pattern (phonetic symbol string) are registered in the voice dictionary database 52 while being sent to 10. By repeating the above operation, the voice dictionary required by the user is automatically created in the voice dictionary database 52. In addition, the minimum number of words, such as various commands, state or province names,
The main / sub category is a voice dictionary database 52
You can register in advance. Also, not only the words input when setting the actual destination, but also the city name input screen is displayed on the screen of the navigation device for the purpose of registration in the voice dictionary database 52 and the predetermined city name is input by voice. By doing so, the city name is converted into a voice dictionary database 52.
You can register at. Similarly, by inputting a predetermined street name by voice on the street name input screen, the street name can be registered in the voice dictionary database 52. Furthermore, similarly, the facility name can be registered in the voice dictionary database 52 by voice inputting a predetermined facility name on the facility name input screen.

FIG. 2 is a block diagram of the navigation device 10. In the figure, 11 is a map storage medium for storing map information, such as a DVD (digital video disc), 12 is a DVD
A DVD control unit for controlling reading of map information from the disc, 13
Is a position measuring device for measuring the current position of the vehicle, 14 is a map information memory for storing the map information around the vehicle position read from the DVD, 15 is various settings and instructions by menu selection, and a point name is input and enlargement / reduction is performed. Reference numeral 16 denotes a remote controller interface for performing operations such as.

Reference numeral 17 is a CPU (navigation control unit) for controlling the entire navigation device, 18 is software (loading program) for downloading various control programs from the DVD, ROM for storing fixed data and the like, and 19 is download from the DVD. It is a RAM that stores various control programs (destination setting program, route search program, etc.) that have been performed, searched route data, and other processing results. Reference numeral 21 denotes a storage unit that stores an address database read from a DVD, in which cities are listed for each state and streets are listed for each city, and the latitude and longitude of a point indicated by an address (state / city / street / address). have.

Reference numeral 22 denotes a storage unit for storing a POI database read from a DVD, which holds an address, a category, a POI position (longitude) and the like for each POI name. 23
Is a display controller for generating a map image / guidance route, 24 is a video RAM for storing images generated by the display controller, 25 is a menu / list generating unit for generating various menus and lists, and 26 is a composite image output. An image synthesizing unit, 27 is a display device (monitor) for displaying the image output from the image synthesizing unit, 28 is a distance to the intersection, a traveling direction, a voice guidance unit for outputting a voice input guidance message by voice, and 29 is communication. The interface is for transmitting and receiving data to and from the voice recognition device 50.

FIG. 3 is a block diagram of the voice recognition apparatus, and the same parts as those in FIG. 1 are designated by the same reference numerals. Speech recognition engine 5
1 includes a voice recognition unit 54 and a voice dictionary creation unit 55. The voice recognition unit 55 uses a voice dictionary database 52 for voice recognition.
Referring to, the pattern matching is performed between the voice pattern of the speaker input from the microphone 56 and the registered pattern, the word with the highest similarity (the highest score) is determined, and the word is sent to the navigation device 10. The voice dictionary creation unit 55 sends voice pattern data (voice waveform data) of a word that is voice input at the time of word registration to the external server 70 for voice recognition via the wireless communication unit 53, and the external server performs voice recognition. The word and the voice pattern of the word are paired and registered in the voice dictionary database 52. The operation unit 57 includes a talk switch TKS for executing voice recognition and a registration mode switch for setting the registration mode.
It has VRS etc.

FIG. 4 is a voice dictionary creation processing flow according to the first embodiment of the present invention. When in the registration mode by operating the registration mode switch VRS (step 201), the talk switch TKS is operated on a predetermined screen of the navigation device, and when there is a voice input thereafter (step 202), the voice dictionary creation unit 55 Sends voice waveform data of the input voice to the external server 70 and requests voice recognition (step 203). The external server 70 performs pattern matching using the received voice waveform data and the phonetic symbols registered in the large-capacity voice dictionary database 80 to search for the word with the highest degree of similarity, and to search for the character string of the word. The phonetic symbol string is returned to the voice dictionary creating unit 55. As a result, the voice dictionary creating unit 55 acquires the character string and the phonetic symbol string (pronunciation pattern) of the recognized word (step 204), and checks whether the word is already registered in the voice dictionary database 52.
(Step 205), if not registered, it is registered in alphabetical order in the voice dictionary database 52 (Step 206), and then the recognized word is sent to the navigation device 10 (Step
207), and the process ends.

In the first embodiment, the correspondence between words and voice patterns is simply registered in the voice dictionary database 52 in alphabetical order. However, the correspondence may be registered separately for each navigation screen. If you do this,
When speech recognition is performed using the speech dictionary of the speech dictionary database 52, the number of search target words is reduced, and high-speed and highly accurate speech recognition is possible.

As shown in FIG. 5 (a), a command dictionary section D1 is provided for all screens, an address dictionary section D2 is provided for address input screens, and a POI dictionary section D3 is provided for facility input screens. As shown in FIG. 5 (b), the address dictionary unit D2 further stores a correspondence between the state name and its phonetic symbol string (pronunciation pattern) in association with the state input screen. Dictionary section D2 that stores the correspondence between city names and their phonetic symbol strings in association with the input screen
2. A dictionary unit D23 is provided for each city to store the correspondence between street names and their phonetic symbol strings in association with the street input screen. Further, as shown in FIG. 5 (c), the POI dictionary unit D3 further corresponds to the category input screen and the dictionary unit D31 that stores the correspondence between the main category name and its phonetic symbol string, and the subcategory input screen. Dictionary section D32, which stores the correspondence between subcategory names and their phonetic symbol strings, for each category,
A dictionary unit D33 is provided for storing the correspondence between the facility name and its phonetic symbol string in association with the facility name input screen.

In other words, the voice dictionary database 52 is provided with various dictionary units for registering a subordinate concept word and its voice pattern as a set for each superordinate concept. For an address,
The superordinate concept and the subordinate concept are a state input screen ID and a state name, a city input screen ID and a city name for each state, and a street input screen ID and a street name for each city. In the case of POI, the above-mentioned superordinate concept and subordinate concept are the category input screen ID and main category name, the subcategory input screen ID and subcategory name for each main category, the facility name input screen and facility for each main or sub category. Is the name. When a word of a subordinate concept is input by voice, the voice dictionary creating unit 55 registers the word obtained by requesting the external server 70 for recognition and its voice pattern in the dictionary unit according to the superordinate concept.

FIG. 6 shows a voice dictionary when the voice dictionary database 52 is provided with a command dictionary section D1, an address dictionary section D2 corresponding to the address input screen, and a POI dictionary section D3 corresponding to the facility name input screen. It is a creation processing flow. It is assumed that all commands are registered in the command dictionary section D1 in advance. When in registration mode by operating the registration mode switch VRS
(Step 301), the talk switch TKS is operated on the predetermined screen of the navigation device, and then, when there is a voice input (Step 302), the voice dictionary creation unit 55 acquires the screen ID from the navigation device 10 (Step 303). .

Next, the voice dictionary recognition section 54 performs voice recognition using the words registered in the command dictionary section D1 and checks whether the input voice is a command (step 304). If it is a command, the command is input to the navigation device 10. The navigation device 10 performs a process according to the input command to create and display a predetermined screen (step 305). After that, the process returns to the beginning and the subsequent processes are repeated. If it is not a command in step 304, the voice recognition unit 54 performs a dictionary search according to the screen ID (step 306) and checks whether a word search has been performed (step 307). If you can search
The recognition word is input to the navigation device 10. The navigation device 10 performs processing according to the input word to create and display a predetermined screen (step 308). After that, the process returns to the beginning and the subsequent processes are repeated.

In step 307, if no search is possible,
The voice dictionary creating unit 55 sends the voice waveform data of the input voice to the external server 70 and requests voice recognition (step 309). The external server 70 performs pattern matching using the received voice waveform data and the phonetic symbols registered in the large-capacity voice dictionary data base 80 to search for the word with the highest degree of similarity,
The character string of the word and the phonetic symbol string are returned to the voice dictionary creating unit 55. As a result, the voice dictionary creating unit 55 acquires the character string of the recognized word and the phonetic symbol string (pronunciation pattern) (step 310), and associates the character string of the recognized word with the phonetic symbol string according to the screen ID. To register (step 311). Then, the voice dictionary creating unit 55 inputs the recognized word to the navigation device 10.
The navigation device 10 performs processing according to the input word to create and display a predetermined screen (step 308). After that, the process returns to the beginning and the subsequent processes are repeated.

As described above, the word can be registered in the voice dictionary database in association with the screen ID, and the word registration and the voice recognition processing can be performed in parallel. In addition, the large-capacity voice dictionary database 80 is registered with words in the same system as the voice dictionary database 52. By doing this, the voice dictionary creation unit 55 can use the screen ID when requesting voice recognition from the external server 70.
Is transmitted together with the voice waveform data, and the external server 70 displays the screen.
Only the words in the dictionary corresponding to the ID are searched for as words. By doing so, the number of words to be searched can be reduced, so that high-speed word recognition can be performed.

FIG. 7 shows a modified example of FIG. 6, in which the voice recognition unit 54 performs only command identification processing, and the same parts as those in FIG. 6 are designated by the same step numbers. In FIG. 7, up to step 305 is the same. If it is not a command in step 304, the voice dictionary creating unit 55 sends the voice waveform data of the input voice to the external server 70 and requests voice recognition (step 401). The external server 70 performs pattern matching using the received voice waveform data and the phonetic symbols registered in the large-capacity voice dictionary database 80 to search for the word with the highest degree of similarity, and to search for the character string of the word. The phonetic symbol string is returned to the voice dictionary creating unit 55. As a result, the voice dictionary creating unit 55 acquires the character string of the recognized word and the phonetic symbol string (phonetic pattern).
(Step 402). Then, the voice dictionary creating unit 55 checks whether or not the recognized word is registered in the voice dictionary database 52 (step 403), and if not, associates the character string of the recognized word with the phonetic symbol string. It is registered in the dictionary unit corresponding to the screen ID (step 404). Then, the voice dictionary creating unit 55 inputs the recognized word to the navigation device 10. The navigation device 10 performs processing according to the input word to create and display a predetermined screen (step 405). After that, the process returns to the beginning and the subsequent processes are repeated.

As described above, in the processing flows of FIGS. 6 and 7, all the command words are registered in advance in the command dictionary section D1 of the voice dictionary database 52. It can also be configured to register based on the voice recognition result from the external server.

FIG. 8 shows a processing flow when a voice dictionary is created in the voice dictionary database 52 and voice recognition is performed using the voice dictionary, and voice registration is not performed. When the user presses the talk switch TKS on the remote control on the map display screen 501, the system makes a beep sound. This "pip"
The generation of sound enables voice input (same below), and the user inputs "Menu" by voice. Since the command dictionary corresponds to all the screens, the voice recognition unit 54 uses the command dictionary unit D.
Voice recognition is performed by referring to 1, and "Menu" is identified and input to the navigation device 10.

The navigation control unit displays "F" for specifying the destination input method on the screen by inputting "Menu".
ind Destination by "Display the screen 502," How would you
Then, the user voice-inputs "Address". Since the command dictionary D1 corresponds to the screen 502, the voice recognition unit 54 refers to the command dictionary unit D1. Voice recognition is performed, and "Address" is identified and input to the navigation device 10.

The navigation control unit displays the menu items "Change State" and "Cit" on the screen by inputting "Address".
"Find address by" with "Y Name" and "Street Name"
Display screen 503 and select "Select City or Street to find yo."
"ur address in California." is output by voice. In addition, since the last selected state is displayed as the default state, if you want to change the state name, input "Change State" by voice, and then change the state. Enter your name by voice.

After the state is determined, the user voice-inputs "City" to input the address of the destination starting from the city name.
Since the command dictionary D1 corresponds to the screen 503, the voice recognition unit 54 recognizes voice by referring to the command dictionary unit D1,
"city" is identified and input to the navigation device 10. The navigation control displays the "Enter City Name" screen 504 with an alphabet / numeric keyboard for entering the city name on the screen by entering "City", Which City in California? "Is output as a voice.
When the user inputs the city name "TORRANCE", the navigation control unit displays the dictionary D22 according to the city input screen of California.
Is referred to for voice recognition, "TORRANCE" is identified and input to the navigation device 10.

The navigation controller has an "Enter St" which has an alphabet / numeric keyboard for entering the city name on the screen by entering "TORRANCE".
Display reetName "Screen 505," then "Which Street in Torranc
e? "Is voice output. The user selects the street name" Carson "
Then, the navigation control unit recognizes voice by referring to the dictionary D23 corresponding to the street input screen of TORRANCE, identifies "Carson", and inputs it to the navigation device 10. After that, the address can be input by voice by the same operation.

[0040]

As described above, according to the present invention, the input voice pattern is sent to the external voice recognition device, and the correspondence between the word recognized by the external voice recognition device and the voice pattern is used in the voice dictionary database. It is highly reliable because it is configured to recognize a voice word input by using a voice dictionary registered in the voice dictionary database registered and registered in the voice dictionary database. , You can create a voice dictionary of only the necessary words.

Further, according to the present invention, a dictionary unit for registering a subordinate concept word and its voice pattern as a set is provided for each superordinate concept, and when the subordinate concept word is input by voice, it is obtained from an external voice recognition device. Since the registered words and their voice patterns are registered in the dictionary section according to the superordinate concept, the number of words to be pattern-matched can be reduced because only the words registered in a predetermined dictionary section need to be searched. , High-speed voice recognition will be possible.

Further, according to the present invention, when a voice is input, the word recognition process of the input voice is performed using the voice dictionary, and when the word can be recognized, the external voice recognition device is not inquired and the word cannot be recognized. Only in the case of inquiry to the external voice recognition device, since the correspondence between the word voice-recognized by the external voice recognition device and the voice pattern of the word is registered in the voice dictionary database, the user is automatically recognized while performing the voice recognition. It is possible to create a voice dictionary in which voice patterns of words required by the are registered.

[Brief description of drawings]

FIG. 1 is an overall configuration diagram of a system that realizes a voice dictionary creating method of the present invention.

FIG. 2 is a configuration diagram of a navigation device.

FIG. 3 is a configuration diagram of a voice recognition device.

FIG. 4 is a voice dictionary creation processing flow according to the first embodiment of the present invention.

FIG. 5 is a storage example of a voice dictionary database.

FIG. 6 is a processing flow of a second embodiment of creating a voice dictionary.

FIG. 7 is a flow chart of a voice dictionary creation process when the voice recognition unit performs only a command identification process.

FIG. 8 is a processing flow for setting a destination by voice inputting an address.

FIG. 9 is an explanatory diagram of a conventional destination specifying method by inputting an address.

FIG. 10 is another explanatory diagram of a conventional destination specifying method by inputting an address.

FIG. 11 is an explanatory diagram of a destination specifying method by inputting a conventional POI type (category).

FIG. 12 is an explanatory diagram in the case of inputting a destination address by voice in the related art.

FIG. 13 is an explanatory diagram of a voice recognition device.

[Explanation of symbols]

10 navigation devices 50 voice recognition device 51 Speech recognition engine 52 Personal voice dictionary database 53 Wireless communication section 70 External server for voice recognition 80 large dictionary database

Claims (7)

[Claims] 1. A voice pattern of a plurality of words is registered, an input voice pattern is compared with a registered voice pattern, and a voice-input word is recognized based on the most similar registered voice pattern. In the voice recognition device,
A voice dictionary database for registering correspondences between words and voice patterns, sending an input voice pattern to an external voice recognition device, and a voice dictionary showing correspondences between words recognized by the external voice recognition device and voice patterns of the words. A voice dictionary creating unit for registering in the database, and a voice recognizing unit for recognizing a word of voice input using the voice dictionary registered in the voice dictionary database are provided. Speech recognizer. 2. The voice dictionary database has a dictionary unit for registering a subordinate concept word and its voice pattern as a set for each superordinate concept, and when the subordinate concept word is input by voice, 2. The voice recognition device according to claim 1, wherein the dictionary creation unit registers the word obtained from the external voice recognition device and its voice pattern in the dictionary unit according to the superordinate concept. 3. When a voice recognition device is used for voice input of a navigation device, said superordinate concept and subordinate concept are state input screen ID and state name, city input screen ID and city name for each state, street for each street. 3. The voice recognition device according to claim 2, wherein the input screen and the street name are used. 4. When the voice recognition device is used for voice input of a navigation device, the superordinate concept and subordinate concept are a category input screen ID and category name, a facility input screen for each category and a facility name. The voice recognition device according to claim 2. 5. When a voice is input, the voice recognition unit performs word recognition processing of the input voice using a voice dictionary, and when the word can be recognized, the voice dictionary creation unit converts the input voice pattern into an external voice. When the word cannot be recognized without being sent to the recognition device, the input voice pattern is sent to the external voice recognition device, and the correspondence between the word recognized by the external voice recognition device and the voice pattern of the sound word is converted into a voice dictionary database. The voice recognition device according to claim 1 or 2, wherein the voice recognition device is registered in the voice recognition device. 6. A voice pattern of a plurality of words is registered, the input voice pattern is compared with the registered voice pattern, and the voice input word is recognized based on the most similar registered voice pattern. In the method for creating a voice dictionary of a voice recognition device, a voice dictionary database for registering correspondence between words and voice patterns is provided, the input voice pattern is sent to an external voice recognition device, and voice recognition is performed by the external voice recognition device. A voice dictionary creating method characterized by creating a voice dictionary by acquiring correspondence between a word and a voice pattern of the word and registering the correspondence in a voice dictionary database. 7. When a voice is input, the voice dictionary is used to perform word recognition processing of the input voice, and when the word can be recognized, the input voice pattern is not sent to an external voice recognition device, and the word is recognized. If it cannot be recognized, the input voice pattern is sent to the external voice recognition device, and the correspondence between the word voice-recognized by the external voice recognition device and the voice pattern of the word is registered in the voice dictionary database to create the voice dictionary. The voice dictionary creating method according to claim 6, wherein the voice dictionary is created.