IP Library Granted Patent US 10,339,918
Granted Patent B2
US 10,339,918 · App. 15/277,164 · Granted Jul 2, 2019

Adaptive speech endpoint detector

Inventors: Joachim Hofer (Munich, DE); Munir Georges (Kehl, DE)
Assignee: Intel IP Corporation
G10L15/04G10L15/07G10L15/22G10L25/87G10L2025/786
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 10,339,918
App. No.
15/277,164
Granted
Jul 2, 2019
Kind
B2
Abstract

An embodiment of a speech endpoint detector apparatus may include a speech detector to detect a presence of speech in an electronic speech signal, a pause duration measurer communicatively coupled to the speech detector to measure a duration of a pause following a period of detected speech, an end of utterance detector communicatively coupled to the pause duration measurer to detect if the pause measured following the period of detected speech is greater than a pause threshold corresponding to an end of an utterance, and a pause threshold adjuster to adaptively adjust the pause threshold corresponding to an end of an utterance based on stored pause information. Other embodiments are disclosed and claimed.

Claims (73)

1. A speech endpoint detector apparatus, comprising:

a speech detector including one or more of configurable logic or fixed functionality logic to detect a presence of speech in an electronic speech signal;

a pause duration measurer including one or more of configurable logic or fixed functionality logic communicatively coupled to the speech detector to measure a duration of a pause following a period of detected speech;

an end of utterance detector including one or more of configurable logic or fixed functionality logic communicatively coupled to the pause duration measurer to detect if the pause measured following the period of detected speech is greater than a pause threshold corresponding to an end of an utterance; and

a pause threshold adjuster including one or more of configurable logic or fixed functionality logic communicatively coupled to the end of utterance detector to store the measured duration of pauses in the detected speech in stored pause information associated with the user and adaptively adjust the pause threshold corresponding to an end of an utterance based on stored pause durations associated with a user, wherein the pause threshold adjuster is further to determine statistics of pause durations, and wherein the stored pause information includes a database having at least two sets of pause statistics respectively associated with at least two identified users.

2. The speech endpoint detector of claim 1 , wherein the stored pause information further includes pause information associated with one or more contextual interpretations.

3. The speech endpoint detector apparatus of claim 2 , wherein the stored pause information associated with the one or more contextual interpretations includes one or more of pause statistics corresponding to one or more phrase contexts, pause information corresponding to an environmental context, or pause information corresponding to a visual context.

4. The speech endpoint detector apparatus of claim 3 , wherein the stored pause information includes pause information associated with the user stored in a finite state transducer.

5. A speech endpoint detector apparatus, comprising:

a speech detector including one or more of configurable logic or fixed functionality logic to detect a presence of speech in an electronic speech signal;

a pause duration measurer including one or more of configurable logic or fixed functionality logic communicatively coupled to the speech detector to measure a duration of a pause following a period of detected speech;

an end of utterance detector including one or more of configurable logic or fixed functionality logic communicatively coupled to the pause duration measurer to detect if the pause measured following the period of detected speech is greater than a pause threshold corresponding to an end of an utterance; and

a pause threshold adjuster including one or more of configurable logic or fixed functionality logic communicatively coupled to the end of utterance detector to adaptively adjust the pause threshold corresponding to an end of an utterance based on stored pause information associated with a user, wherein the stored pause information further includes pause information associated with one or more contextual interpretations, and

wherein the stored pause information associated with the one or more contextual interpretations includes one or more of pause statistics corresponding to one or more phrase contexts, pause information corresponding to an environmental context, or pause information corresponding to a visual context, and

wherein the stored pause information includes pause information associated with the user stored in a finite state transducer.

6. The speech endpoint detector apparatus of claim 5 , wherein the pause threshold adjuster is further to determine statistics of pause durations, and wherein the stored pause information associated with the user includes a database of pause statistics associated with an identified user.

7. The speech endpoint detector apparatus of claim 6 , wherein the stored pause information includes a database having at least two sets of pause statistics respectively associated with at least two identified users.

8. A method of detecting an endpoint of speech, comprising:

detecting a presence of speech in an electronic speech signal;

measuring a duration of a pause following a period of detected speech;

detecting if the pause measured following the period of detected speech is greater than a pause threshold corresponding to an end of an utterance;

adaptively adjusting the pause threshold corresponding to an end of an utterance based on stored pause information associated with a user;

storing the measured duration of pauses in the detected speech in the stored pause information associated with the user;

adjusting the pause threshold based on the stored pause durations associated with the user;

determining statistics of pause durations associated with an identified user; and

storing a database having at least two sets of pause statistics respectively associated with at least two identified users.

9. The method of detecting an endpoint of speech of claim 8 , wherein the stored pause information further includes pause information associated with one or more contextual interpretations.

10. The method of detecting an endpoint of speech of claim 9 , further comprising:

determining statistics of pause durations associated with one or more phrase contexts; and

storing a database of pause statistics associated with the one or more phrase contexts in the stored pause information.

11. The method of detecting an endpoint of speech of claim 10 , further comprising:

storing pause information associated with the user in a finite state transducer.

12. A method of detecting an endpoint of speech comprising:

detecting a presence of speech in an electronic speech signal;

measuring a duration of a pause following a period of detected speech;

detecting if the pause measured following the period of detected speech is greater than a pause threshold corresponding to an end of an utterance;

adaptively adjusting the pause threshold corresponding to an end of an utterance based on stored pause information associated with a user, wherein the stored pause information further includes pause information associated with one or more contextual interpretations;

determining statistics of pause durations associated with one or more phrase contexts;

storing a database of pause statistics associated with the one or more phrase contexts in the stored pause information; and

storing pause information associated with the user in a finite state transducer.

13. The method of detecting an endpoint of speech of claim 12 , further comprising:

determining statistics of pause durations associated with an identified user; and

storing a database of pause statistics associated with the identified user in the stored pause information associated with the identified user.

14. The method of detecting an endpoint of speech of claim 13 , further comprising:

storing a database having at least two sets of pause statistics respectively associated with at least two identified users.

15. At least one non-transitory computer readable medium comprising a set of instructions, which when executed by a computing device, cause the computing device to:

detect a presence of speech in an electronic speech signal;

measure a duration of a pause following a period of detected speech;

detect if the pause measured following the period of detected speech is greater than a pause threshold corresponding to an end of an utterance;

adaptively adjust the pause threshold corresponding to an end of an utterance based on stored pause information associated with a user;

store the measured duration of pauses in the detected speech in the stored pause information associated with the user;

adjust the pause threshold based on the stored pause durations associated with the user;

determine statistics of pause durations associated with an identified user; and

store a database having at least two sets of pause statistics respectively associated with at least two identified users.

16. The at least one non-transitory computer readable medium of claim 15 , wherein the stored pause information further includes pause information associated with one or more contextual interpretations.

17. The at least one non-transitory computer readable medium of claim 15 , comprising a further set of instructions, which when executed by a computing device, cause the computing device to:

determine statistics of pause durations associated with one or more phrase contexts; and

store a database of pause statistics associated with the one or more phrase contexts in the stored pause information.

18. The at least one non-transitory computer readable medium of claim 17 , comprising a further set of instructions, which when executed by a computing device, cause the computing device to:

store pause information associated with the user in a finite state transducer.

19. At least one non-transitory computer readable medium of claim 15 , comprising a set of instructions, which when executed by a computing device, cause the computing device to:

detect a presence of speech in an electronic speech signal;

measure a duration of a pause following a period of detected speech;

detect if the pause measured following the period of detected speech is greater than a pause threshold corresponding to an end of an utterance;

adaptively adjust the pause threshold corresponding to an end of an utterance based on stored pause information associated with a user, wherein the stored pause information further includes pause information associated with one or more contextual interpretations;

determine statistics of pause durations associated with one or more phrase contexts;

store a database of pause statistics associated with the one or more phrase contexts in the stored pause information; and

store pause information associated with the user in a finite state transducer.

20. The at least one non-transitory computer readable medium of claim 19 , comprising a further set of instructions, which when executed by a computing device, cause the computing device to:

determine statistics of pause durations associated with an identified user; and

store a database of pause statistics associated with the identified user in the stored pause information associated with the identified user.

21. The at least one non-transitory computer readable medium of claim 20 , comprising a further set of instructions, which when executed by a computing device, cause the computing device to:

store a database having at least two sets of pause statistics respectively associated with at least two identified users.

Assignments (2)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded May 22, 2021
From: INTEL IP CORPORATION
To: INTEL CORPORATION
Reel/Frame 056337/0609 →
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Oct 21, 2016
From: HOFER, JOACHIM; GEORGES, MUNIR
To: INTEL IP CORPORATION
Reel/Frame 040466/0470 →
Continuity (1)
Related Publication 20180090127A1 · Mar 29, 2018
Cited By (13)
US 12,322,381 US 12,380,889 US 12,400,661 US 12,406,672 US 12,406,684 US 12,412,567 US 12,456,465 US 12,462,808 US 12,494,929 US 12,518,748 US 12,555,581 US 12,561,289 US 12,586,576