Cascading Specialized Recognition Engines Based on a Recognition Policy

• US 20180025731A1
• Filed: 07/21/2016
• Published: 01/25/2018
• Est. Priority Date: 07/21/2016
• Status: Abandoned Application

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Inventors: Lovitt, Andrew
• Application Number:
  US15/216,576
• Publication Number:
  US 20180025731A1
• Time in Patent Office: Days
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 15/285 Memory allocation or algorithm optimisation to reduce hardware requirements
• 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 17/10 Multimodal systems, i.e. based on the integration of multiple recognition engines or fusion of expert systems
• CPC Class Codes: G10L 2015/088 Word spotting

Abstract

Specialized recognition engines are configured to recognize acoustic objects. A policy engine can consume a recognition policy that defines the conditions under which specialized recognition engines are to be activated or deactivated. An arbitrator receives events fired by the specialized recognition engines and provides the events to listeners that have registered to receive notification of the occurrence of the events. If a specialized recognition engine recognizes an acoustic object, the policy engine can utilize the recognition policy to identify the specialized recognition engines that are to be activated or deactivated. The identified specialized recognition engines can then be activated or deactivated in order to implement a particular recognition scenario and to meet a particular power consumption requirement.

Claims

1. A computer-implemented method, comprising:

activating a first specialized recognition engine on a device, the first specialized recognition engine configured to recognize a first acoustic object on the device;

determining that the first specialized recognition engine has recognized the first acoustic object;

responsive to determining that the first specialized recognition engine has recognized the first acoustic object, selecting a second specialized recognition engine configured to recognize a second acoustic object on the device based upon a recognition policy;

activating the selected second specialized recognition engine on the device;

determining that the device is entering a low power state; and

deactivating the second specialized recognition engine on the device to reduce power consumption in response to determining that the device is entering the low power state.

2. The computer-implemented method of claim 1, further comprising modifying one or more recognition thresholds associated with the first specialized recognition engine responsive to determining that the first specialized recognition engine has recognized the first acoustic object.

3. The computer-implemented method of claim 1, further comprising deactivating the first specialized recognition engine responsive to activating the selected second specialized recognition engine.

4. The computer-implemented method of claim 1, further comprising providing contents of an audio buffer to the second specialized recognition engine.

5. The computer-implemented method of claim 1, further comprising providing contents of an audio buffer to a program registered to receive a notification that the first specialized recognition engine has recognized the first acoustic object.

6. The computer-implemented method of claim 1, wherein the device comprises a digital signal processor (DSP) and a system on a chip (SOC), wherein the first specialized recognition engine executes on the DSP, and wherein a program registered to receive a notification that the first specialized recognition engine has recognized the first acoustic object executes on the SOC.

7. The computer-implemented method of claim 1, further comprising activating a third specialized recognition engine based upon the recognition policy.

8. The computer-implemented method of claim 1, wherein the second specialized recognition engine is selected from a plurality of specialized recognition engines based upon the recognition policy.

10. The computer-implemented method of claim 1, further comprising:

determining that the device is exiting the low power state; and

reactivating the second specialized recognition engine responsive to determining that the device is exiting the low power state.

21. The computer-implemented method of claim 1, further comprising determining that the second specialized recognition engine has recognized the second acoustic object prior to deactivating the second specialized recognition engine.

9. (canceled)

11. An apparatus, comprising:

one or more processors; and

at least one computer storage medium having computer executable instructions stored thereon which, when executed by the one or more processors, cause the apparatus to;

execute a first specialized recognition engine,

execute a policy engine,

receive an indication from the first specialized recognition engine that a first acoustic object has been recognized,

responsive to receiving the indication, select a second specialized recognition engine based upon a recognition policy of the policy engine,

execute the selected second specialized recognition engine to recognize a second acoustic object,

determine that the apparatus is entering a low power state; and

deactivate the second specialized recognition engine to reduce power consumption in response to determining that the apparatus is entering the low power state.

12. The apparatus of claim 11, wherein the at least one computer storage medium has further computer executable instructions stored thereon to:

execute an arbitrator based on the indication received from the first specialized recognition engine that the first acoustic object has been recognized, and

provide a notification to a program registered to receive the notification that the first specialized recognition engine has recognized the first acoustic object.

13. The apparatus of claim 11, wherein the at least one computer storage medium has further computer executable instructions stored thereon to modify one or more recognition thresholds associated with the first specialized recognition engine.

14. The apparatus of claim 11, wherein the at least one computer storage medium has further computer executable instructions stored thereon to deactivate the first specialized recognition engine.

15. The apparatus of claim 11, wherein the at least one computer storage medium has further computer executable instructions stored thereon to provide contents of an audio buffer to the second specialized recognition engine.

16. A computer storage medium having computer executable instructions stored thereon which, when executed on a computing device, cause the computing device to:

activate a first specialized recognition engine configured to recognize a first acoustic object on the computing device;

receive an indication that the first specialized recognition engine has recognized the first acoustic object;

select a second specialized recognition engine configured to recognize a second acoustic object based upon a recognition policy responsive to receiving the indication that the first specialized recognition engine has recognized the first acoustic object;

activate the selected second specialized recognition engine on the computing device;

determine that the computing device is entering a low power state; and

deactivate the second specialized recognition engine on the computing device to reduce power consumption in response to determining that the computing device is entering the low power state.

17. The computer storage medium of claim 16, having further computer executable instructions stored thereon to modify one or more recognition thresholds associated with the first specialized recognition engine.

18. The computer storage medium of claim 16, having further computer executable instructions stored thereon to deactivate the first specialized recognition engine.

19. The computer storage medium of claim 16, having further computer executable instructions stored thereon to provide contents of an audio buffer to the second specialized recognition engine.

20. The computer storage medium of claim 16, having further computer executable instructions stored thereon to activate a third specialized recognition engine configured to recognize a third acoustic object on the computing device.