Weight Coefficient Generation Device, Voice Recognition Device, Navigation Device, Vehicle, Weight Coefficient Generation Method, and Weight Coefficient Generation Program

• US 20110231191A1
• Filed: 11/17/2009
• Published: 09/22/2011
• Est. Priority Date: 11/25/2008
• Status: Active Grant

• Current Assignee: Asahi Kasei Kabushiki Kaisha
• Original Assignee: Asahi Kasei Kabushiki Kaisha
• Inventors: Miyazaki, Toshiyuki
• Granted Patent: US 8,688,449 B2
• Time in Patent Office: Days
• US Class Current: 704/243
• CPC Class Codes: G01C 21/3608 using speech input, e.g. using speech recognition
• CPC Class Codes: G10L 2015/085 Methods for reducing search complexity, pruning

Abstract

A weight coefficient generation device, a speech recognition device, a navigation system, a vehicle, a vehicle coefficient generation method, and a weight coefficient generation program are provided for the purpose of improving a speech recognition performance of place names. In order to address the above purpose, an address database 12 has address information data items including country names, city names, street names, and house numbers, and manages the address information having a tree structure indicating hierarchical relationships between the place names from wide area to a narrow area. Each of the place names stored in the address database 12 is taken as a speech recognition candidate. A weight coefficient calculation unit 11 of a weight coefficient generation device 10 calculates a weight coefficient of the likelihood of the aforementioned recognition candidate based on the number of the street names belonging to the lower hierarchy below the city names.

Claims

1-15. -15. (canceled)

16. A speech recognition device comprising:

a speech recognition dictionary in which a place name for use in an address having a hierarchical structure from a wide area to a narrow area is registered as a vocabulary to be recognized;

a recognition likelihood calculation unit for extracting features from a speech, and calculating likelihood of the speech to each of the vocabularies, to be recognized, registered in the speech recognition dictionary based upon the features;

an address database in which a tree structure indicating a hierarchical relationship of the place names of the vocabularies to be recognized;

a weight coefficient calculation unit for calculating a weight coefficient of likelihood of the vocabulary to be recognized based on an information quantity belonging to lower hierarchy below the vocabulary to be recognized; and

a recognition result output unit for outputting a speech recognition result based on weighted likelihood weighted by multiplying the likelihood of the speech calculated by the recognition likelihood calculation unit and the weight coefficient of the likelihood of the vocabulary to be recognized calculated by the weight coefficient calculation unit.

17. The speech recognition device according to claim 16, wherein:

the address database stores city names and street names belonging to the lower hierarchy below the city names, and

the information quantity belonging to the lower hierarchy below the vocabulary to be recognized in the address database is a number of the street names belonging to the lower hierarchy below the city names.

18. The speech recognition device according to claim 16, wherein:

the address database stores the city names and facility names belonging to the lower hierarchy below the city names, and

the information quantity belonging to the lower hierarchy below the vocabulary to be recognized in the address database is a number of the facility names belonging to the lower hierarchy below the city names.

19. The speech recognition device according to claim 16, wherein:

the address database stores the city names and facility names belonging to the lower hierarchy below the city names, and

the information quantity belonging to the lower hierarchy below the vocabulary to be recognized in the address database is a number of telephone numbers of the facility names belonging to the lower hierarchy below the city names.

20. A weight coefficient generation device that takes each of place names stored in an address database for managing address information having a tree structure indicating a hierarchical relationship of the place names, as a vocabulary to be recognized for a speech uttered by a user, and that generates a weight coefficient of likelihood of the vocabulary to be recognized for the speech, the weight coefficient generation device comprising a weight coefficient calculation unit for calculating the weight coefficient of the likelihood of the vocabulary to be recognized, on the basis of an information quantity belonging to a lower hierarchy below the vocabulary to be recognized in the address database.

21. The weight coefficient generation device according to claim 20, wherein the information quantity belonging to the lower hierarchy below the vocabulary to be recognized in the address database is the number of the place names.

22. The weight coefficient generation device according to claim 21, wherein the address database stores city names and street names belonging to the lower hierarchy below the city names, and

wherein the weight coefficient calculation unit calculates the weight coefficient of the likelihood of the city names on the basis of the number of the street names belonging to the lower hierarchy below the city names.

23. The weight coefficient generation device according to claim 20, wherein the information quantity belonging to the lower hierarchy below the vocabulary to be recognized in the address database is the number of facilities.

24. The weight coefficient generation device according to claim 23, wherein the address database stores city names and facility names belonging to the lower hierarchy below the city names, and

wherein the weight coefficient calculation unit calculates the weight coefficient of the likelihood of the city names on the basis of the number of the facility names belonging to the lower hierarchy below the city names.

25. The weight coefficient generation device according to claim 20, wherein the information quantity belonging to the lower hierarchy below the vocabulary to be recognized in the address database is the number of telephone numbers of the facilities.

26. The weight coefficient generation device according claim 25, wherein the address database stores the city names or the facility names, and the telephone numbers belonging to the lower hierarchy below the city names or the facility names, and

wherein the weight coefficient calculation unit calculates the weight coefficient of the likelihood of the city names or the facility names on the basis of the number of the telephone numbers belonging to the lower hierarchy below the city names or the facility names.

27. The weight coefficient generation device according claim 20, wherein the information quantity belonging to the lower hierarchy below the vocabulary to be recognized in the address database is a used storage capacity by the lower hierarchy below the vocabulary to be recognized.

28. The weight coefficient generation device according to one of claim 20 to claim 27, further comprising an information quantity output unit for calculating the information quantity belonging to the lower hierarchy below the vocabulary to be recognized in the address database.

29. A speech recognition device recognizing a place name uttered by a user, the device comprising:

a speech input unit for allowing the user to input a speech;

a speech recognition dictionary in which the place names used for addresses having a hierarchical structure from a wide area to a narrow area are registered as a vocabulary to be recognized; and

a recognition result output unit for calculating likelihood of the inputted speech for the vocabulary to be recognized stored in the speech recognition dictionary,

wherein a weight coefficient generated by the weight coefficient generation device of one of claims 20 to 27 further registered in the speech recognition dictionary, and

wherein the recognition result output unit outputs a speech recognition result on the basis of a weighted likelihood which is multiplication of the weight coefficient by the calculated likelihood.

30. The speech recognition device according to claim 29, wherein the recognition result output unit selects a plurality of weighted likelihood each having a high value from among the weighted likelihood, and displays a plurality of vocabularies to be recognized corresponding to each of the selected plurality of weighted likelihood, in order to determine one vocabulary to be recognized by user'"'"'s selective input.

31. A weight coefficient generation method that takes each of place names stored in an address database for managing address information having a tree structure indicating a hierarchical relationship of the place names, as a vocabulary to be recognized for a speech uttered by a user, and that generates a weight coefficient of likelihood of the vocabulary to be recognized for the speech, the weight coefficient generation method comprising calculating the weight coefficient of the likelihood of the vocabulary to be recognized, on the basis of an information quantity belonging to a lower hierarchy below the vocabulary to be recognized in the address database.

32. A weight coefficient generation program that takes each of place names stored in an address database for managing address information having a tree structure indicating a hierarchical relationship of the place names, as a vocabulary to be recognized for a speech uttered by a user, and that generates a weight coefficient of likelihood of the vocabulary to be recognized for the speech, the program causing a computer to execute a weight coefficient generation step of calculating the weight coefficient of the likelihood of the vocabulary to be recognized, on the basis of an information quantity belonging to a lower hierarchy below the vocabulary to be recognized in the address database.