Speech recognition using automatic recognition turn off

• US 20050043954A1
• Filed: 09/24/2004
• Published: 02/24/2005
• Est. Priority Date: 09/05/2001
• Status: Active Grant

• Current Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Original Assignee: Voice Signal Technologies Incorporated (Microsoft Corporation)
• Inventors: Roth, Daniel L., Johnston, David F., Cohen, Jordan R.
• Granted Patent: US 7,716,058 B2
• Time in Patent Office: Days
• US Class Current: 704/275
• CPC Class Codes: G06F 3/165 Management of the audio stream, e.g. setting of volume, audio stream path
• CPC Class Codes: G10L 15/19 Grammatical context, e.g. disambiguation of the recognition hypotheses based on word sequence rules
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue

Abstract

Large vocabulary speech recognition can automatically turn recognition off in one or more ways. A user command can turn on recognition that is automatically turned off after the next end of utterance. A plurality of buttons can each be associated with a different speech mode and the touch of a given button can turn on, and then automatically turn off, the given button'"'"'s associated speech recognition mode. These selectable modes can include large vocabulary and alphabetic entry modes, or continuous and discrete modes. A first user input can start recognition that allows a sequence of vocabulary words to be recognized and a second user input can start recognition that turns off after one word has been recognized. A first user input can start recognition that allows a sequence of utterances to be recognized and a second user input can start recognition that allows only a single utterance to be recognized.

Claims

1. A computerized method of performing speech recognition comprising:

responding to a command input from a user to start recognition by;

turning large vocabulary speech recognition on after the receipt of the command;

subsequently automatically turning the large vocabulary speech recognition off and leaving it off until receiving another command input from a user to start the recognition;

wherein the turning off a speech recognition occurs automatically in response to the first end of utterance detected after said turning on of the speech recognition.

2. A method as in claim 1 wherein the command input which causes the turning on of speech recognition is a non-acoustic input.

3. A method as in claim 2 wherein the speech recognition is continuous speech recognition.

4. A method as in claim 2 wherein the speech recognition is discrete speech recognition.

5. A method as in claim 2 further comprising:

outputting a user perceivable representation of the one or more words recognized as a best choice for the utterance preceding the end of utterance detection;

then providing a user interface allowing a user to provide correction input to correct errors in the best choice output in response to the recognition of an utterance; and

then responding to another of said start recognition command inputs generated after the outputting of the best choice recognized for an utterance before any correction input has been received for said best choice by;

confirming said best choice as correct; and

repeating said method again for a new utterance starting with receipt of the start recognition command.

6. A method as in claim 5 further including responding to such a confirmation of a best choice by including one or more of the recognized words in said best choice as being part of the current language context used to calculate a language model score for use in subsequent speech recognition.

7. A method as in claim 5 further include responding to such a confirmation of a best choice by using one or more of the recognized words in the best choice as data for altering a language model.

8. A method as in claim 5 further including responding to such a confirmation of a best choice containing a given recognized word by labeling acoustic data from the best choice'"'"'s corresponding utterance for use in updating one or more acoustic models used in the recognition in of the given recognized word.

9. A method as in claim 2 further including allowing a user to select between a first mode in which recognition turns off after the next end of utterance that is detected after receiving the non-acoustic input, and second mode which does not turn off recognition after said next end of utterance detection.

10. A method as in claim 9 wherein, in said second mode, recognition is automatically turned off in response to a lapse of time of multiple seconds.

11. A method as in claim 2 wherein:

the method is performed by software running on a handheld computing device; and

the non-acoustic input is the pressing of a button this is either a hardware button or a software created GUI button.

12. A method as in claim 11 wherein

the handheld computing device is a cellphone;

and

the buttons are cellphone buttons.

13. A method as in claim 2 wherein the method is performed by software running on a computer which is part of an automotive vehicle.

14. A method as in claim 2 wherein:

said method provides a user interface having a plurality of speech mode selection buttons, which are either hardware or software buttons, each for selecting a respective one from a plurality of different speech recognition modes, which are distinguished from one another by characteristics other than recognition duration and which are all available for selection by the user at one time; and

the non-acoustic input which causes the turning on of speech recognition is the pressing of one of said buttons; and

the method responds to the pressing of a given speech mode button by turning on speech recognition in the given button'"'"'s associated mode and then subsequently automatically turning off said recognition in said mode.

15. A method as in claim 14 wherein:

the speech recognition mode associated with one of said buttons is a mode that performs said large vocabulary recognition; and

the recognition mode associated with another of said buttons is an alphabetic entry mode which performs recognition with a vocabulary for alphabetic entry.

16. A method as in claim 14 wherein:

the speech recognition mode associated with one of said buttons is a large vocabulary continuous recognition mode; and

the recognition mode associated with another of said buttons is a large vocabulary discrete recognition mode.

17. A method as in claim 14 wherein

the handheld computing device is a cellphone;

and

the buttons are cellphone buttons.

18. A method as in claim 2 wherein:

said method responds to a touch of a given button as said start recognition command input from the user and responds to such a touch lasting less than a first duration as a click, and to such a touch lasting longer than a second duration as a press;

said method responds to a press by causing speech recognition to be performed on sound for a duration that varies as a function of the length of the press; and

said method responds to a click by causing speech recognition to be performed on sound until said first end of utterance after said click is detected.

19. A method as in claim 18 wherein recognition performed in response to a click is discrete recognition and recognition performed in response to a press is continuous recognition.

20. A method as in claim 19 wherein the user interface allows a user to select between:

a mode in which recognition in response to a click and recognition in response to a press are both either continuous or discrete; and

a mode wherein recognition performed in response to a click is discrete recognition and recognition performed in response to a press is continuous recognition.

21. A method as in claim 18 wherein:

the method is practiced on a cellphone; and

numbered cellphone buttons act as said mode buttons.

22. A computerized method of performing speech recognition comprising:

providing a user interface having a plurality of speech mode selection buttons, each being either a hardware or software button that is associated with a respective one from a plurality of different speech recognition modes, which modes are distinguished from one another by characteristics other than recognition duration, and which buttons are all available for selection by the user at one time; and

responding to the pressing of a given one of said speech mode buttons by turning on speech recognition in the given button'"'"'s associated mode and then subsequently automatically turning off said recognition;

wherein;

the speech recognition mode associated with one of said buttons is a mode that performs large vocabulary recognition; and

the recognition mode associated with another of said buttons is an alphabetic entry mode which performs recognition with a vocabulary for alphabetic entry.

23. A method as in claim 22 wherein:

said method responds to a touch of a given speech mode selection button lasting less than a first duration as a click, and to such a touch lasting longer than a second duration as a press;

said method responds to a press by causing speech recognition in the given button'"'"'s associated mode to be performed on sound for a duration that varies as a function of the length of the press; and

said method responds to a click by causing speech recognition in the given button'"'"'s associated mode to be performed on sound for a duration that is independent of the particular duration of the touch that has been responded to as said click.

24. A method as in claim 22 wherein:

the method is practiced on a cellphone; and

numbered cellphone buttons act as said speech mode selection buttons.

25. A computerized method of performing speech recognition comprising:

providing a user interface having a plurality of speech mode selection buttons, each being either a hardware or software button that is associated with a respective one from a plurality of different speech recognition modes, which modes are distinguished from one another by characteristics other than recognition duration, and which buttons are available for selection by the user at one time; and

responding to the pressing of a given one of said speech mode buttons by turning on speech recognition in the given button'"'"'s associated mode and then subsequently automatically turning off said recognition;

wherein;

the speech recognition mode associated with one of said buttons is a mode that performs discrete speech recognition; and

the recognition mode associated with another of said buttons is a mode which performs continuous speech recognition.

26. A method as in claim 25 wherein:

said method responds to a touch of a given speech mode selection button lasting less than a first duration as a click, and to such a touch lasting longer than a second duration as a press;

said method responds to a press by causing speech recognition in the given button'"'"'s associated mode to be performed on sound for a duration that varies as a function of the length of the press; and

said method responds to a click by causing speech recognition in the given button'"'"'s associated mode to be performed on sound for a duration that is independent of the particular duration of the touch that has been responded to as said click.

27. A method as in claim 25 wherein:

the method is practiced on a cellphone; and

numbered cellphone buttons act as said speech mode selection buttons.

28. A computerized method of performing speech recognition comprising:

responding to a first user input by starting a first speech recognition mode that allows a sequence of vocabulary words to be recognized in response to said first input; and

responding to a second user input by starting a second speech recognition mode that allows only one vocabulary words to be recognized in response to said second input.

29. A computerized method of performing speech recognition comprising:

responding to a first user input by starting a first speech recognition mode that allows a sequence of utterances to be recognized in response to said first input; and

responding to a second user input by starting a second speech recognition mode that allows only one utterance to be recognized in response to said second input.