Method and apparatus for voice controlled devices with improved phrase storage, use, conversion, transfer, and recognition

• US 7,283,964 B1
• Filed: 05/09/2000
• Issued: 10/16/2007
• Est. Priority Date: 05/21/1999
• Status: Expired due to Fees

• Current Assignee: Winbond Electronics Corporation
• Original Assignee: Winbond Electronics Corporation
• Inventors: Pyo, Phillip Paul, Barel, Avraham, Brown, Amos, Bootsma, Karin Lissette, Gaddy, Lawrence Kent, MacMillan, David, Geilhufe, Michael
• Primary Examiner(s): Opsasnick, Michael N
• Application Number:
  US09/567,858
• Time in Patent Office: 2,716 Days
• Field of Search: 704/270, 704/251, 704/255, 704/277
• US Class Current: 704/277
• CPC Class Codes: G10L 15/30 Distributed recognition, e.g. in client-server systems, for mobile phones or network applications
• CPC Class Codes: G10L 15/32 Multiple recognisers used in sequence or in parallel; Score combination systems therefor, e.g. voting systems
• CPC Class Codes: G10L 2015/025 Phonemes, fenemes or fenones being the recognition units

Abstract

The embodiments of the invention provide for the storage of speech phrases. Speech phrases are processed by a speaker-independent speech recognition engine of a voice controlled device. This engine returns a speaker-independent representation of the phrase. The speaker-independent representation is stored. Embodiments of the invention include methods of converting text to speaker-independent representations of speech and speaker-independent representations of speech into text.

Claims

1. A method of transferring speaker independent representations of at least one speaker dependent spoken word between voice controlled devices, the method comprising:

receiving at least one speaker dependent spoken word at a first voice controlled device having a first speaker dependent spoken word recognizer;

converting at the first voice controlled device, the at least one speaker dependent spoken word into at least one speaker independent representation of the at least one speaker dependent spoken word; and

,

transferring the at least one speaker independent representation of the at least one speaker dependent spoken word to a second voice controlled device for recognizing the speaker independent representation of speech, the second voice controlled device having a second speaker dependent spoken word recognizer for recognizing the at least one speaker dependent spoken word at the second voice controlled device, and converting at the second voice controlled device, the at least one speaker dependent spoken word into at least one speaker independent representation of the at least one speaker dependent spoken word.

2. The method of claim 1, wherein,

the at least one speaker independent representation of the at least one speaker independent spoken word is a phonetic representation of the at least one speaker dependent spoken word.

3. The method of claim 1, wherein,

the first voice controlled device utilizes one type of speaker independent representation of the at least one speaker dependent spoken word and the second voice controlled device uses a different type of speaker independent representation of the at least one speaker dependent spoken word and the method further comprises:

converting the at least one speaker independent representation of the at least one speaker dependent spoken word from the first voice controlled device into a type of speaker independent representation of the at least one speaker dependent spoken word compatible with the second voice controlled device.

4. A method of transferring speaker independent representations of speaker dependent spoken words between voice controlled devices, the method comprising:

receiving at least one speaker dependent spoken word at a first voice controlled device having a first the at least one speaker dependent spoken word recognizer;

converting the at least one speaker dependent spoken word into at least one speaker independent representation of the at least one speaker dependent spoken word;

providing a set of speaker independent representations of the at least one speaker dependent spoken word associated with a second voice controlled device having a second at least one speaker dependent spoken word recognizer; and

transferring the at least one speaker independent representation of the at least one speaker dependent spoken word from the first voice controlled device to the second voice controlled device for combining with the set of speaker independent representations of the at least one speaker dependent spoken word, the second voice controlled device recognizing at least one speaker dependent spoken word at the second voice controlled device, and for recognizing a speaker independent representation of speech.

5. The method of claim 4, wherein,

the set of speaker independent representations of the at least one speaker dependent spoken word is empty.

6. The method of claim 4, wherein,

the set of speaker independent representations of the at least one speaker dependent spoken word has one speaker independent representation of the at least one speaker dependent spoken word.

7. The method of claim 4, wherein,

the set of speaker independent representations of the at least one speaker dependent spoken word has more than one speaker independent representation of the at least one speaker dependent spoken word.

8. The method of claim 4, wherein,

the at least one speaker independent representation of the at least one speaker dependent spoken word is a phonetic representation of the at least one speaker dependent spoken word.

9. The method of claim 4, wherein,

the at least one speaker independent representation of the at least one speaker dependent spoken word of the first voice controlled device transferred to the second voice controlled device is combined with the set of speaker independent representations of the at least one speaker dependent spoken word in the second voice controlled device by merging the at least one speaker independent representation of the at least one speaker dependent spoken word with the second set of speaker independent representations of the at least one speaker dependent spoken word.

10. The method of claim 4, wherein,

the at least one speaker independent representation of the at least one speaker dependent spoken word of the first voice controlled device transferred to the second voice controlled device is combined with the set of speaker independent representations of the at least one speaker dependent spoken word in the second voice controlled device by replacing the set of speaker independent representations of the at least one speaker dependent spoken word in its entirety with the at least one speaker independent representation of the at least one speaker dependent spoken word.

11. The method of claim 4, wherein,

the at least one speaker independent representation of the at least one speaker dependent spoken word of the first voice controlled device transferred to the second voice controlled device is combined with the set of speaker independent representations of the at least one speaker dependent spoken word in the second voice controlled device by selectively replacing elements of the set of speaker independent representations with elements of the at least one independent representation of the at least one speaker dependent spoken word.

12. The method of claim 4, wherein,

there are date stamps associated with the at least one speaker independent representation of the at least one speaker dependent spoken word and with the set of speaker independent representations of the at least one speaker dependent spoken word, and the at least one speaker independent representation of the at least one speaker dependent spoken word is combined with the second set speaker independent representations of the at least one speaker dependent spoken word using the date stamps.

13. The method of claim 4, wherein,

the first the at least one speaker dependent spoken word recognizer of the first voice controlled device is a first speaker independent the at least one speaker dependent spoken word recognizer;

the second the at least one speaker dependent spoken word recognizer of the second voice controlled device is a second speaker independent the at least one sneaker dependent spoken word recognizer; and

the first speaker independent the at least one speaker dependent spoken word recognizer is of the same design as the second speaker independent the at least one speaker dependent spoken word recognizer.

14. The method of claim 4, wherein,

the first the at least one speaker dependent spoken word recognizer of the first voice controlled device is a first speaker independent the at least one speaker dependent spoken word recognizer;

the second the at least one speaker dependent spoken word recognizer of the second voice controlled device is a second speaker independent the at least one speaker dependent spoken word recognizer; and

the first speaker independent the at least one speaker dependent spoken word recognizer is of a different design than the second speaker independent the at least one speaker dependent spoken word recognizer.

15. The method of claim 4, wherein,

the first voice controlled device and the second voice controlled device use different speaker independent representations of the at least one sneaker dependent spoken word, and

the method further comprises:

converting the at least one speaker independent representation of the at least one speaker dependent spoken word into a type of speaker independent representation of the at least one speaker dependent spoken word compatible with the second voice controlled device.

16. The method of claim 15, wherein,

the type of speaker independent representation of the at least one speaker dependent spoken word compatible with the second voice controlled device is a phonetic representation of the at least one speaker dependent spoken word.

17. A method of providing interoperability for speaker independent the at least one speaker dependent spoken word recognizers, the method comprising:

providing a first the at least one speaker dependent spoken word recognizer;

the first the at least one speaker dependent spoken word recognizer receiving at least one speaker dependent spoken word;

converting the at least one speaker dependent spoken word into at least one speaker independent representation of the at least one speaker dependent spoken word;

providing the at least one speaker independent representation of the at least one speaker dependent spoken word to a second the at least one speaker dependent spoken word recognizer, the second speech recognizer recognizing at least one speaker dependent spoken word at the second voice controlled device, and for recognizing a speaker independent representation of speech;

wherein the first the at least one speaker dependent spoken word recognizer and the second the at least one speaker dependent spoken word recognizer use different types of speaker independent representations of the at least one speaker dependent spoken word; and

the method further comprises;

converting the speaker independent representation of the at least one speaker dependent spoken word of the first the at least one speaker dependent spoken word recognizer into a type of speaker independent representation of the at least one speaker dependent spoken word compatible with the second the at least one sneaker dependent spoken word recognizer.

18. The method of claim 17, wherein,

the first the at least one sneaker dependent spoken word recognizer is within a first device and the second the at least one speaker dependent spoken word recognizer is within a second device.

19. The method of claim 17, wherein,

the first the at least one speaker dependent spoken word recognizer and the second the at least one speaker dependent spoken word recognizer are the same design.

20. The method of claim 17, wherein,

the first the at least one speaker dependent spoken word recognizer and the second the at least one speaker dependent spoken word recognizer are of different designs.

21. The method of claim 17, wherein,

the speaker independent representations of the at least one speaker dependent spoken word are phonetic representations of the at least one speaker dependent spoken word.

22. The method of claim 17, wherein,

the at least one speaker independent representation of the at least one speaker dependent spoken word is a phonetic representation of the at least one speaker dependent spoken word.

23. The method of claim 17, wherein,

the type of speaker independent representation of the at least one speaker dependent spoken word compatible with the second the at least one speaker dependent spoken word recognizer is a phonetic representation of the at least one speaker dependent spoken word.

24. A method of transferring speaker independent representations of speaker dependent spoken words between voice controlled devices, the method comprising:

providing a first the at least one speaker dependent spoken word recognizer that operates with a first type of speaker independent representation of the at least one speaker dependent spoken word;

the first the at least one speaker dependent spoken word recognizer receiving at least one speaker dependent spoken word;

converting the at least one speaker dependent spoken word into at least one speaker independent representation;

providing a second the at least one speaker dependent spoken word recognizer that operates with a second type of speaker independent representation of the at least one speaker dependent spoken word, the second speech recognizer recognizing at least one speaker dependent spoken word at the second voice controlled device, and for recognizing the second type of speaker independent representation of speech;

providing a conversion means for converting the first type of speaker independent representation of the at least one speaker dependent spoken word into a type compatible with the second the at least one speaker dependent spoken word recognizer; and

providing a conversion means for converting the second type of speaker independent representation of the at least one speaker dependent spoken word into a type compatible with the first the at least one speaker dependent spoken word recognizer.

25. The method of claim 24, wherein,

at least one of the first or second type of speaker independent representations of the at least one speaker dependent spoken word is a phonetic representation of the at least one speaker dependent spoken word.

26. A method for converting the at least one speaker dependent spoken word into text, the method comprising:

receiving at least one speaker dependent spoken word of one or more spoken words;

converting the at least one speaker dependent spoken word into at least one speaker independent representation of the at least one speaker dependent spoken word;

communicating at least one second speaker dependent spoken word;

converting the at least one second speaker dependent spoken word into at least one speaker independent representation of the at least one speaker dependent spoken word for recognition;

comparing the at least one speaker independent representation of the at least one speaker dependent spoken word for recognition with the at least one speaker independent representation of the at least one speaker dependent spoken word to generate a match result; and

converting the at least one speaker independent representation of the at least one speaker dependent spoken word into text.

27. The method of claim 26, further comprising:

storing the at least one speaker independent representation of the at least one speaker dependent spoken word in a storage.

28. The method of claim 26, wherein,

the converting is performed by a the at least one speaker dependent spoken word-to-text converter.

29. The method of claim 26, wherein,

the converting provides one or more possible text results and the method further comprises:

processing at least one of the possible text results to create at least one possible speaker independent representation of the at least one speaker dependent spoken word; and

comparing the at least one speaker independent representation of the at least one speaker dependent spoken word stored in storage with the at least one possible speaker independent representation of the at least one speaker dependent spoken word to generate text.

30. The method of claim 26, wherein,

the at least one speaker independent representation of the at least one speaker dependent spoken word is a phonetic representation of the at least one speaker dependent spoken word.

31. The method of claim 26, further comprising:

providing a first additional data associated with the speaker independent representations of the at least one speaker dependent spoken word stored in storage;

providing one or more possible text results;

providing a second additional data associated with the possible text results; and

using relationships between the first additional data and the second additional data to assist in the converting.

32. The method of claim 31, wherein,

the first additional data and the second additional data include phone number information.

33. The method of claim 31, wherein,

the converting includes having a user speak the spelling of at least part of the text.

34. A method for developing the at least one speaker dependent spoken word recognizers that can use stored speaker dependent spoken words from other the at least one speaker dependent spoken word recognizers, the method comprising:

identifying stored speaker independent representations of the at least one speaker dependent spoken word associated with a first speaker independent the at least one speaker dependent spoken word recognizer, where at least some of the stored speaker independent representations of the at least one speaker dependent spoken word are created by providing at least one speaker dependent spoken word of speaker dependent representations of spoken words to the first speaker independent the at least one speaker dependent spoken word recognizer; and

providing a second speaker independent the at least one sneaker dependent spoken word recognizer that can use the stored speaker independent representations of the at least one speaker dependent spoken word associated with the first speaker independent the at least one speaker dependent spoken word recognizer; and

,

communicating the stored speaker independent representations of speech from the first speaker independent speech recognizer to the second speaker independent speech recognizer for use in recognizing speaker dependent representation of speech by the second speaker independent speech recognizer.

35. The method of claim 34, wherein,

the first speaker independent the at least one speaker dependent spoken word recognizer uses a first type of speaker independent representation of the at least one speaker dependent spoken word;

the second speaker independent the at least one speaker dependent spoken word recognizer uses a second type of speaker independent representation of the at least one speaker dependent spoken word; and

to recognize speaker dependent spoken words, the second speaker independent the at least one speaker dependent spoken word recognizer uses the stored speaker independent representations of the at least one speaker dependent spoken word associated with the first speaker independent the at least one speaker dependent spoken word recognizer by converting the stored speaker independent representations of the at least one speaker dependent spoken word associated with the first speaker independent the at least one speaker dependent spoken word recognizer into a type compatible with the second speaker independent the at least one speaker dependent spoken word recognizer.

36. A method for developing the at least one speaker dependent spoken word recognizers that can provide stored speaker dependent spoken words to other the at least one speaker dependent spoken word recognizers, the method comprising:

providing a first speaker independent the at least one speaker dependent spoken word recognizer that creates speaker independent representations of the at least one speaker dependent spoken word from speaker dependent spoken words; and

communicating speaker independent representations of speech from the first speaker independent speech recognizer to a second speaker independent the at least one speaker dependent spoken word recognizer for use of the speaker independent representations of the at least one speaker dependent spoken word for recognizing speaker dependent spoken words.

37. The method of claim 36, wherein,

the first speaker independent the at least one speaker dependent spoken word recognizer uses a first type of speaker independent representation of the at least one speaker dependent spoken word;

the second speaker independent the at least one speaker dependent spoken word recognizer uses a second type of speaker independent representation of the at least one speaker dependent spoken word; and

to recognize speaker dependent spoken words, the second speaker independent the at least one speaker dependent spoken word recognizer uses the stored speaker independent representations of the at least one speaker dependent spoken word associated with the first speaker independent the at least one speaker dependent spoken word recognizer by converting the stored speaker independent representations of the at least one speaker dependent spoken word associated with the first speaker independent the at least one speaker dependent spoken word recognizer into a type compatible with the second speaker independent the at least one speaker dependent spoken word recognizer.