Method and system for considering information about an expected response when performing speech recognition

• US 8,756,059 B2
• Filed: 12/30/2010
• Issued: 06/17/2014
• Est. Priority Date: 02/04/2005
• Status: Active Grant

• Current Assignee: Vocollect, Inc. (Honeywell International Inc.)
• Original Assignee: Vocollect, Inc. (Honeywell International Inc.)
• Inventors: Braho, Keith, El-Jaroudi, Amro, Pike, Jeffrey
• Primary Examiner(s): MCFADDEN, SUSAN IRIS
• Application Number:
  US12/982,563
• Publication Number:
  US 20110093269A1
• Time in Patent Office: 1,265 Days
• Field of Search: 704/251
• US Class Current: 704/251
• CPC Class Codes: G10L 15/08 Speech classification or search
• CPC Class Codes: G10L 15/1815 Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning

Abstract

A speech recognition system receives and analyzes speech input from a user in order to recognize and accept a response from the user. Under certain conditions, information about the response expected from the user may be available. In these situations, the available information about the expected response is used to modify the behavior of the speech recognition system by taking this information into account. The modified behavior of the speech recognition system comprises adjusting the rejection threshold when speech input matches the predetermined expected response.

Claims

1. A method for recognizing speech, the method comprising the steps of:

analyzing speech input with an apparatus using at least one hardware-implemented processor to generate a hypothesis including multiple elements;

generating respective confidence factors for the multiple elements of the speech input;

comparing a confidence factor for at least one of the elements to a threshold for accepting the element as properly recognized; and

if the element is not accepted in the comparison, comparing at least a portion of the hypothesis containing the unaccepted element to at least a portion of an expected response;

if the comparison of the portion of the hypothesis containing the unaccepted element with the portion of the expected response is favorable, storing the unaccepted element for subsequent analysis and proceeding to further analyze another element of the hypothesis.

2. The method of claim 1 wherein the elements include one or more words that make up the speech input.

3. The method of claim 1 wherein each element is an individual word.

4. The method of claim 1 further comprising, after storing at least one unaccepted element for subsequent analysis, adjusting the acceptance threshold and comparing the confidence factor for at least one stored element to the adjusted acceptance threshold.

5. The method of claim 1 further comprising, if the comparison of the portion of the hypothesis containing the unaccepted element with the portion of the expected response is not favorable, rejecting at least a portion of the hypothesis.

6. The method of claim 5 further comprising proceeding to analyze elements of the hypothesis until the full hypothesis is analyzed or a portion of the hypothesis is rejected.

7. The method of claim 1 further comprising the step of either increasing or decreasing the acceptance threshold based on the favorable comparison of the hypothesis portion to the expected response.

8. The method of claim 1, further comprising comparing the confidence factors of elements in the hypothesis to an acceptance threshold in a sequential fashion.

9. The method of claim 4 wherein an adjustment amount for the acceptance threshold is a fixed amount.

10. The method of claim 4 wherein an adjustment amount for the acceptance threshold is a variable amount.

11. The method of claim 1 further comprising generating the hypothesis in an element-by-element fashion, and the subsequent analysis for a stored element being performed after generating the next element in the hypothesis.

12. A system for recognizing speech, comprising:

a hardware-implemented processor that has a speech recognizer configured to analyze speech input and generate a hypothesis of multiple elements and respective confidence factors for the multiple elements;

an acceptance algorithm that compares the confidence factor of at least one element to an acceptance threshold for accepting or not accepting the element;

an adjustment component that compares at least a portion of the hypothesis containing an element that is not accepted to at least a portion of an expected response; and

if the comparison of the portion of the hypothesis containing the unaccepted element with the portion of the expected response is favorable, the system operable for storing the unaccepted element for subsequent analysis and proceeding to further analyze another element of the hypothesis.

13. The system of claim 12 wherein the elements include one or more words that make up the speech input.

14. The system of claim 12 wherein each element is an individual word.

15. The system of claim 12 wherein the adjustment component, after at least one unaccepted element is stored for subsequent analysis, adjusts the acceptance threshold and compares the confidence factor for at least one stored element to the adjusted acceptance threshold.

16. The system of claim 12 wherein the system rejects at least a portion of the hypothesis if the comparison of the portion of the hypothesis containing the unaccepted element to the portion of the expected response is not favorable.

17. The system of claim 12 wherein the system generates the hypothesis in an element-by-element fashion, and the subsequent analysis for a stored element is performed after the speech recognizer generates the next element in the hypothesis.

18. A method for recognizing speech, the method comprising the steps of:

analyzing speech input with an apparatus using at least one hardware-implemented processor to generate a hypothesis of multiple elements and respective confidence factors for the multiple elements;

comparing each said confidence factor to a threshold for accepting the element; and

comparing at least a portion of the hypothesis to at least a portion of an expected response;

if the comparison is favorable, adjusting a threshold for one or more of the elements in order to affect the acceptance of the one or more elements.