System and method for speech recognition using a reduced user dictionary, and computer readable storage medium therefor

• US 8,676,582 B2
• Filed: 03/14/2008
• Issued: 03/18/2014
• Est. Priority Date: 03/14/2007
• Status: Active Grant

• Current Assignee: NEC Corporation
• Original Assignee: NEC Corporation
• Inventors: Yamabana, Kiyoshi
• Primary Examiner(s): He, Jialong
• Application Number:
  US12/526,465
• Publication Number:
  US 20100324899A1
• Time in Patent Office: 2,195 Days
• Field of Search: 704/251, 704/270
• US Class Current: 704/251
• CPC Class Codes: G10L 15/07 to the speaker

Abstract

A speech recognition system for rapidly performing recognition processing while maintaining quality of speech recognition in a speech recognition device, are provided. A speech recognition system includes a speech input device which inputs speech and displays a recognition result, and a speech recognition device which receives the speech from the speech input device, performs recognition processing, and sends back the speech to the speech input device. The speech input device includes a user dictionary section which stores words used for recognizing the input speech, and a reduced user dictionary creation unit which extracts words corresponding to the input speech from the user dictionary and creates a reduced user dictionary. The speech recognition device has a speech recognition unit which inputs the input speech and the reduced user dictionary from the speech input/output device and recognizes the input speech based on the reduced user dictionary and a system dictionary provided beforehand.

Claims

1. A speech recognition system, comprising:

a speech input device configured to input a speech as an input speech by converting the speech to an electric signal; and

a speech recognition device configured to take in and recognize the speech input to the speech input device,

wherein the speech input device and the speech recognition device are communicatively connected to each other,

wherein the speech input device comprises;

a user dictionary section storing words for use with recognition of the input speech; and

a reduced user dictionary creation unit configured to create a reduced user dictionary by extracting words corresponding to the input speech from the user dictionary section, and

wherein the speech recognition device comprises;

a speech recognition unit configured to

input the input speech and the reduced user dictionary from the speech input device and

recognize the input speech based on the reduced user dictionary and a built-in system dictionary storing words for speech recognition,

wherein the reduced user dictionary creation unit includes;

a comparing section which compares the input speech and a word in the user dictionary and compiles a likelihood of each word appearing in the input speech,

a word temporarily storing section which temporarily stores a set of each word and a corresponding likelihood compiled, and

a word selection section which selects one or a plurality of words having high usage from the word temporarily storing section to thereby create the reduced user dictionary.

2. The speech recognition system as claimed in claim 1, wherein the reduced user dictionary is a dictionary in which a word having a possibility of being included in the input speech is selected from the words in the user dictionary.

3. The speech recognition system as claimed in claim 2, wherein the reduced user dictionary creation unit compares a word in the user dictionary and the input speech, calculates a likelihood that the word appears in the input speech, and based on a calculation result, selects a word having a high likelihood to thereby create the reduced user dictionary.

4. The speech recognition system as claimed in claim 2, wherein the reduced user dictionary creation unit creates the reduced user dictionary by means of a word spotting method using the user dictionary.

5. The speech recognition system as claimed in claim 1, wherein the reduced user dictionary creation unit compares a word in the user dictionary and the input speech, calculates a likelihood that the word appears in the input speech, and based on a calculation result, selects a word having a high likelihood to thereby create the reduced user dictionary.

6. The speech recognition system as claimed in claim 1, wherein the reduced user dictionary creation unit creates the reduced user dictionary by means of a word spotting method using the user dictionary.

7. A speech recognition system, comprising:

a speech input device configured to input a speech as an input speech by converting the speech to an electric signal; and

a speech recognition device configured to take in and recognize the input speech input to the speech input device,

wherein the speech input device and the speech recognition device are communicatively connected to each other,

wherein the speech input device comprises;

a speech input section for inputting a speech;

a user dictionary section storing words for use with recognition of the input speech;

a reduced user dictionary creation unit configured to create a reduced user dictionary by extracting words corresponding to the input speech from the user dictionary; and

a transmission unit configured to transmit the input speech and the reduced user dictionary to the speech recognition device, and

wherein the speech recognition device comprises;

a system dictionary section storing words for speech recognition;

a receiving unit configured to receive the input speech and the reduced user dictionary transmitted from the speech input device; and

a speech recognition unit configured to perform speech recognition on the input speech by using the system dictionary and the reduced user dictionary,

wherein the reduced user dictionary creation unit includes;

a comparing section which compares the input speech and a word in the user dictionary and compiles a likelihood of each word appearing in the input speech,

a word temporarily storing section which temporarily stores a set of each word and a corresponding likelihood compiled, and

a word selection section which selects one or a plurality of words having high usage from the word temporarily storing section to thereby create the reduced user dictionary.

8. The speech recognition system as claimed in claim 7, wherein the reduced user dictionary is a dictionary in which a word having a possibility of being included in the input speech is selected from the words in the user dictionary.

9. The speech recognition system as claimed in claim 7, wherein the reduced user dictionary creation unit compares a word in the user dictionary and the input speech, calculates a likelihood that the word appears in the input speech, and based on a calculation result, selects a word having a high likelihood to thereby create the reduced user dictionary.

10. The speech recognition system as claimed in claim 7, wherein the reduced user dictionary creation unit creates the reduced user dictionary by means of a word spotting method using the user dictionary.

11. A speech recognition method, comprising:

converting a speech into an electric signal and inputting the converted speech as an input speech into a speech input device;

extracting words related to the input speech from a user dictionary for speech recognition included in the speech input device to create a reduced user dictionary;

transmitting the input speech and the reduced user dictionary from the speech input device to a speech recognition device; and

at the speech recognition device having received the input speech and the reduced user dictionary, performing speech recognition on the input speech based on a system dictionary for speech recognition included in the speech recognition device and the received reduced user dictionary,

wherein the extracting includes;

comparing the input speech and a word in the user dictionary and compiling a likelihood of each word appearing in the input speech,

temporarily storing a set of each word and a corresponding likelihood compiled, and

selecting one or a plurality of words having high usage to thereby create the reduced user dictionary.

12. The speech recognition method as claimed in claim 11, wherein the reduced user dictionary is created by:

comparing words included in the user dictionary with the input speech,

calculating likelihood of each word being present in the input speech, and

selecting a word having high likelihood of being present in the input speech based on a result of the calculation.

13. The speech recognition method as claimed in claim 11, wherein the reduced user dictionary is created from the user dictionary by word spotting.

14. The speech recognition method as claimed in claim 11, wherein the reduced user dictionary is created by:

comparing the input speech with words in the user dictionary,

calculating likelihood of each word being present in the input speech,

temporarily storing a set of the calculated likelihood and a corresponding word,

selecting one or more words having high likelihood from temporarily-stored words, and

editing the selected words to be in a form of a dictionary.

15. A speech recognition method, comprising:

converting a speech into an electric signal and inputting the converted speech as an input speech into a speech input device;

extracting words related to the input speech from a user dictionary for speech recognition included in the speech input device to create a reduced user dictionary;

transmitting the input speech and the reduced user dictionary from the speech input device to a speech recognition device;

receiving at the speech recognition device the input speech and the reduced user dictionary; and

at the speech recognition device, performing speech recognition on the input speech based on a system dictionary for speech recognition included in the speech recognition device and the received reduced user dictionary,

wherein the extracting includes;

comparing the input speech and a word in the user dictionary and compiling a likelihood of each word appearing in the input speech,

temporarily storing a set of each word and a corresponding likelihood compiled, and

selecting one or a plurality of words having high usage to thereby create the reduced user dictionary.

16. The speech recognition method as claimed in claim 15, wherein the reduced user dictionary is created by:

comparing words included in the user dictionary with the input speech,

calculating likelihood of each word being present in the input speech, and

selecting a word having high likelihood of being present in the input speech based on a result of the calculation.

17. The speech recognition method as claimed in claim 15, wherein the reduced user dictionary is created from the user dictionary by word spotting.

18. The speech recognition method as claimed in claim 15, wherein the reduced user dictionary is created by:

comparing the input speech with words in the user dictionary,

calculating likelihood of each word being present in the input speech,

temporarily storing a set of the calculated likelihood and a corresponding word,

selecting one or more words having high likelihood from temporarily-stored words, and

editing the selected words to be in a form of a dictionary.

19. A non-transitory computer readable storage medium storing a speech recognition program, in a speech recognition system including a speech input device configured to input a speech as an input speech by converting the speech to an electric signal and a speech recognition device configured to take in and recognize the speech input to the speech input device, for causing a computer included in the speech input device to execute:

converting speech into an electric signal and inputting the converted speech as input speech into the speech input device;

extracting words related to the input speech from a user dictionary for speech recognition included in the speech input device to create a reduced user dictionary; and

transmitting the input speech and the reduced user dictionary from the speech input device to a speech recognition device for performing speech recognition,

wherein the extracting includes;

comparing the input speech and a word in the user dictionary and compiling a likelihood of each word appearing in the input speech,

temporarily storing a set of each word and a corresponding likelihood compiled, and

selecting one or a plurality of words having high usage to thereby create the reduced user dictionary.

20. The non-transitory computer readable storage medium storing a speech recognition program as claimed in claim 19, wherein the reduced user dictionary is created by:

comparing words included in the user dictionary with the input speech,

calculating likelihood of each word being present in the input speech, and

selecting a word having high likelihood of being present in the input speech based on a result of the calculation.

21. The non-transitory computer readable storage medium storing a speech recognition program as claimed in claim 19, wherein the reduced user dictionary is created by:

comparing the input speech with words in the user dictionary,

compiling likelihood of each word being present in the input speech,

temporarily storing a set of the compiled likelihood and a corresponding word, and

selecting one or more words having high likelihood from temporarily-stored words.

22. The non-transitory computer readable storage medium storing a speech recognition program as claimed in claim 19, wherein the reduced user dictionary is created from the user dictionary by word spotting.

23. A speech recognition system, comprising:

speech input means for inputting a speech as an input speech by converting the speech to an electric signal; and

speech recognition means for taking in and recognizing the speech input to the speech input device,

wherein the speech input means and the speech recognition means are communicatively connected to each other,

wherein the speech input means comprises;

user dictionary means for storing words for use with recognition of the input speech; and

reduced user dictionary creation means for creating a reduced user dictionary by extracting words corresponding to the input speech from the user dictionary section, and

wherein the speech recognition means comprises;

speech recognition means for

inputting the input speech and the reduced user dictionary from the speech input means and

recognizing the input speech based on the reduced user dictionary and a built-in system dictionary storing words for speech recognition,

wherein the reduced user dictionary creation means includes;

a comparing section which compares the input speech and a word in the user dictionary and compiles a likelihood of each word appearing in the input speech,

a word temporarily storing section which temporarily stores a set of each word and a corresponding likelihood compiled, and

a word selection section which selects one or a plurality of words having high usage from the word temporarily storing section to thereby create the reduced user dictionary.