AGGREGATE CONTEXT INFERENCES USING MULTIPLE CONTEXT STREAMS

• US 20130297547A1
• Filed: 03/13/2013
• Published: 11/07/2013
• Est. Priority Date: 05/07/2012
• Status: Active Grant

• Current Assignee: Qualcomm, Inc.
• Original Assignee: Qualcomm, Inc.
• Inventors: Ding, Li, Narayanan, Vidya, Lee, Jin Won, Kuhn, Lukas D., Grokop, Leonard Henry, Sadasivam, Shankar, Manolakos, Alexandros
• Granted Patent: US 9,582,755 B2
• Time in Patent Office: Days
• US Class Current: 706/46
• CPC Class Codes: G06N 5/02 Knowledge representation; Symbolic representation
• CPC Class Codes: H04M 1/72448 with means for adapting the functionality of the device according to specific conditions
• CPC Class Codes: H04M 2250/12 including a sensor for measuring a physical value, e.g. temperature or motion

Abstract

Methods, systems, computer-readable media, and apparatuses for inferring context are provided. In one potential implementation, first context information associated with a first duration is identified, second context information is accessed to determine a context segmentation boundary; and the first context information and the second context information is then aggregated to generate an inferred segmented aggregated context. In a further implementation, the first context information is used to average inferred contexts, and the context segmentation boundary is used to reset a start time for averaging the first context information.

Claims

1. A method comprising:

accessing first context information associated with a first duration;

accessing second context information to determine a context segmentation boundary; and

aggregating the first context information and the second context information to generate an inferred segmented aggregated context.

2. The method of claim 1 wherein the first context information is ambient sound from a microphone and wherein the second context information is speech information from the microphone.

3. The method of claim 1, further comprising obtaining the first context information associated with the first duration at a mobile device using one or more sensors;

and

obtaining the second context information using second sensors different from the one or more sensors, wherein the second sensors are at the mobile device or a different device.

4. The method of claim 1, wherein aggregating the first context information and the second context information to generate the inferred segmented aggregated context comprises resetting an average inferred context to begin an average context starting at a time associated with the context segmentation boundary.

5. The method of claim 4 wherein the time associated with the context segmentation boundary comprises a time at which the context segmentation boundary was detected;

and

wherein a mobile device context setting associated with the inferred segmented aggregated context is delayed for a predetermined time following identification of the context segmentation boundary.

6. The method of claim 1, wherein aggregating the first context information and the second context information to generate the inferred segmented aggregated context comprises:

associating the first context information with a first potential context;

determining that the second context information is inconsistent with the first potential context; and

identifying the inferred segmented aggregated context, the inferred segmented aggregated context being consistent with the second context information and different from the first potential context.

7. The method of claim 6, wherein accessing the second context information to determine the context segmentation boundary comprises computing individual context inferences of limited temporal relevance to identify expected context transitions.

8. A method comprising:

determining, by a first computing device, a first set of context inferences covering a first time period, wherein the first set of context inferences is determined from sensor data;

determining from the first set of context inferences, a context segmentation boundary associated with a first time delineating a first context prior to the first time from a second context following the first time within the first time period;

determining, by the first computing device, a second context inference from the sensor data associated with a second time following the first time within the first time period;

determining an inferred segmented aggregated context from the second context and the second context inference; and

selecting a set of device settings for the first computing device based on the inferred segmented aggregated context.

9. The method of claim 8 further comprising identifying the context segmentation boundary from calendar data prior to determining the context segmentation boundary from the first context information.

10. The method of claim 8 further comprising identifying the context segmentation boundary from social networking check-in data prior to determining the context segmentation boundary from the first context information.

11. The method of claim 8 wherein the sensor data comprises microphone data detected based on an audio environment of the first computing device.

12. The method of claim 8 wherein the sensor data comprises location data detected based on a location of the first computing device.

13. The method of claim 8 wherein the sensor data comprises Bluetooth identifiers associated with Bluetooth devices proximate to the first computing device.

14. The method of claim 8 wherein the sensor data comprises accelerometer data and wherein the context segmentation boundary is detected based on a change in the accelerometer data.

15. The method of claim 8 wherein the inferred segmented aggregated context is computed from the first set of context inferences and the second context inference by selecting a most commonly occurring individual context inference over a second time period, wherein the second time period is a subset of the first time period.

16. The method of claim 8 further comprising:

determining a confidence value associated with the inferred segmented aggregated context; and

determining that the confidence value is above a confidence threshold prior to selecting the set of device settings for the first computing device based on the inferred segmented aggregated context.

17. The method of claim 16 wherein the inferred segmented aggregated context is computed from individual context inferences of the first set of context inferences and the second context inference by averaging a likelihood or a log-likelihood associated with each potential context, and outputting a highest value context as the inferred segmented aggregated context.

18. The method of claim 17 wherein the confidence value is computed as a function of averaged likelihood or averaged log-likelihood values for each potential context.

19. A mobile device comprising:

a processor;

a first sensor coupled to the processor; and

a non-transitory storage medium coupled to the processor, the storage medium including instructions that when executed by the processor cause the mobile device to;

access first context information associated with a first duration from the first sensor;

access second context information to determine a context segmentation boundary; and

aggregate the first context information and the second context information to generate an inferred segmented aggregated context.

20. The mobile device of claim 19 wherein the first sensor comprises a microphone and wherein the first context information comprises sound information recorded by the microphone;

and

wherein the second context information comprises a meeting reminder stored in a calendar within the storage medium as part of a calendar application operated by the processor.

21. A device comprising:

means for determining a first set of context inferences covering a first time period, wherein the first set of context inferences is determined from sensor data;

means for determining from the first set of context inferences, a context segmentation boundary associated with a first time delineating a first context prior to the first time from a second context following the first time within the first time period;

means for determining a second context inference from the sensor data associated with a second time following the first time within the first time period;

means for determining an inferred segmented aggregated context from the second context and the second context inference; and

means for selecting a set of device settings based on the inferred segmented aggregated context.

22. The device of claim 21 further comprising:

means for identifying the context segmentation boundary from calendar data prior to determining the context segmentation boundary from the first context information.

23. The device of claim 21 further comprising:

means for determining a confidence value associated with the inferred segmented aggregated context; and

means for determining that the confidence value is above a confidence threshold prior to selecting the set of device settings based on the inferred segmented aggregated context, wherein the inferred segmented aggregated context is computed from individual context inferences of the first set of context inferences and the second context inference by averaging a likelihood or a log-likelihood associated with each potential context, and outputting a highest value context as the inferred segmented aggregated context.

24. A non-transitory computer readable medium comprising computer readable instructions that, when executed by a processor, cause a device to implement a method comprising:

determining, by the device, a first set of context inferences covering a first time period, wherein the first set of context inferences is determined from sensor data;

determining from the first set of context inferences, a context segmentation boundary associated with a first time delineating a first context prior to the first time from a second context following the first time within the first time period;

determining, by the device, a second context inference from the sensor data associated with a second time following the first time within the first time period;

determining an inferred segmented aggregated context from the second context and the second context inference; and

selecting a set of device settings for based on the inferred segmented aggregated context.

25. The non-transitory computer readable medium of claim 24, wherein the sensor data comprises device location data;

and

wherein determining from the first set of context inferences, the context segmentation boundary associated with the first time delineating the first context prior to the first time from the second context following the first time within the first time period comprises identifying a movement across a door threshold to enter a room.

26. The non-transitory computer readable medium of claim 25, wherein the inferred segmented aggregated context comprises a meeting context associated with the room.