Selecting feature types to extract based on pre-classification of sensor measurements

• US 8,775,128 B2
• Filed: 07/10/2013
• Issued: 07/08/2014
• Est. Priority Date: 11/07/2012
• Status: Active Grant

• Current Assignee: Samsung Electronics Co. Ltd.
• Original Assignee: Sensor Platforms, Incorporated (Knowles Corporation)
• Inventors: Meduna, Deborah, Waite, Tom, Rajnarayan, Dev
• Primary Examiner(s): SUGLO, JANET L
• Application Number:
  US13/939,126
• Publication Number:
  US 20140129178A1
• Time in Patent Office: 363 Days
• Field of Search: 702/179, 702/182, 702/183, 702187-189
• US Class Current: 702/189
• CPC Class Codes: A61B 5/1123 Discriminating type of movement, e.g. walking or running A61B5/1116, A61B5/112 take precedence
• CPC Class Codes: G06F 17/00 Digital computing or data processing equipment or methods, specially adapted for specific functions information retrieval, database structures or file system structures therefor G06F16/00
• CPC Class Codes: G06F 3/011 Arrangements for interaction with the human body, e.g. for user immersion in virtual reality blind teaching G09B21/00
• CPC Class Codes: G06F 3/017 Gesture based interaction, e.g. based on a set of recognized hand gestures interaction based on gestures traced on a digitiser G06F3/04883
• CPC Class Codes: G06F 3/0346 with detection of the device orientation or free movement in a 3D space, e.g. 3D mice, 6-DOF [six degrees of freedom] pointers using gyroscopes, accelerometers or tilt-sensors
• CPC Class Codes: G06V 40/20 Movements or behaviour, e.g. gesture recognition recognition of facial expressions G06V40/16
• CPC Class Codes: G06V 40/23 Recognition of whole body movements, e.g. for sport training

Abstract

A processing apparatus including one or more processors and memory receives sensor measurements generated by one or more sensors of one or more devices, pre-classifies the sensor measurements as belonging to one of a plurality of pre-classifications, and selects one or more feature types to extract from the sensor measurements based at least in part on the pre-classification of the sensor measurements. The processing apparatus also extracts features of the one or more selected feature types from the sensor measurements and determines a state of a respective device of the one or more devices in accordance with a classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements.

Claims

1. A method comprising:

at a processing apparatus having one or more processors and memory storing one or more programs that, when executed by the one or more processors, cause the respective processing apparatus to perform the method;

receiving sensor measurements generated by one or more sensors of one or more devices;

pre-classifying the sensor measurements as belonging to one of a plurality of pre-classifications;

selecting one or more feature types to extract from the sensor measurements based at least in part on the pre-classification of the sensor measurements, including;

in accordance with a determination that the sensor measurements correspond to a first pre-classification, selecting a first set of feature types as the one or more selected feature types; and

in accordance with a determination that the set of sensor measurements correspond to a second pre-classification, selecting a second set of feature types as the one or more selected feature types, wherein the first set of feature types is different from the second set of feature types;

extracting features of the one or more selected feature types from the sensor measurements, including;

in accordance with a determination that the sensor measurements correspond to the first pre-classification, forgoing extracting features of one or more of the second set of feature types while extracting features of the first set of feature types; and

in accordance with a determination that the set of sensor measurements correspond to the second pre-classification, forgoing extracting features of one or more of the first set of feature types while extracting features of the second set of feature types; and

determining a state of a respective device of the one or more devices in accordance with a classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements.

2. The method of claim 1, wherein the state of the respective device comprises one or more of:

a physical property of the respective device;

a state of an environment of the respective device; and

a state of an entity associated with the respective device.

3. The method of claim 1, wherein:

pre-classifying the sensor measurements includes identifying the sensor measurements as belonging to a respective pre-classification that is associated with multiple feature sets, including a first set of features for use in identifying states of the respective device under a first set of conditions and a second set of features for use in identifying states of the respective device under a second set of conditions different from the first set of conditions;

extracting the features includes extracting the first set of features and extracting the second set of features; and

determining the state of the respective device includes determining the state of the respective device in accordance with the first set of features and the second set of features.

4. The method of claim 1, wherein:

pre-classifying the sensor measurements includes identifying the sensor measurements as belonging to one of a plurality of coupling pre-classifications; and

determining the state of the respective device includes determining a coupling state of the respective device in accordance with a coupling classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements.

5. The method of claim 4, wherein the coupling state of the respective device is selected from a set consisting of:

a plurality of coupling-stable states corresponding to a coupling between the respective device and an entity associated with the respective device; and

a plurality of coupling-transition states corresponding to a transition between two of the coupling-stable states.

6. The method of claim 1, wherein:

the first set of feature types includes features adapted for determining a state of the respective device while the respective device is stationary; and

the second set of feature types includes features adapted for determining a state of the respective device while the respective device is in motion.

7. The method of claim 1, wherein the first set of feature types are feature types that enable classification of a coupling state of the respective device.

8. The method of claim 1, wherein the second set of feature types are feature types that enable classification of a transition between two different coupling states of the respective device.

9. The method of claim 1, wherein the first set of feature types and the second set of feature types are mutually exclusive.

10. The method of claim 1, wherein the plurality of pre-classifications include:

a first pre-classification corresponding to a plurality of transition feature types associated with identifying a transition between two different states of the respective device;

a second pre-classification corresponding to a first subset of a plurality of device-state feature types associated with identifying a state of the respective device; and

a third pre-classification corresponding to a second subset of the plurality of device-state feature types, wherein the second subset of device-state feature types is different from the first subset of device-state feature types.

11. The method of claim 1, wherein:

the method includes obtaining context information indicative of a current context of the respective device; and

the sensor measurements are pre-classified based at least in part on the current context of the respective device.

12. The method of claim 1, wherein:

pre-classifying the sensor measurements includes extracting features of one or more pre-classification feature types from the sensor measurements; and

extracting features of the pre-classification feature types from the sensor measurements is more resource efficient than extracting features of the one or more selected feature types from the sensor measurements.

13. The method of claim 1, wherein:

the one or more selected feature types includes a respective feature type that provides an approximation of frequency bandwidth information for a respective sensor-measurement signal that corresponds to sensor measurements collected over time; and

extracting features of the one or more selected feature types from the sensor measurements includes extracting features of the respective feature type from the respective sensor-measurement signal in the time domain without converting the respective sensor-measurement signal into the frequency-domain.

14. The method of claim 1, wherein:

determining the state of the respective device includes updating a Markov model, wherein the Markov model includes;

a plurality of model states corresponding to states of the respective device; and

a plurality of sets of transition probabilities between the plurality of model states;

in accordance with a determination that the sensor measurements correspond to a first pre-classification, updating one or more model states in the Markov model in accordance with a first set of transition probabilities; and

in accordance with a determination that the set of sensor measurements correspond to a second pre-classification, updating one or more model states in the Markov model in accordance with a second set of transition probabilities different from the first set of transition probabilities.

15. The method of claim 1, wherein the receiving sensor measurements, pre-classifying sensor measurements, selecting one or more feature types, extracting features and determining the state of the respective device are repeated for a plurality of measurement epochs.

16. The method of claim 15, wherein a state of the respective device determined in a prior measurement epoch is used as a factor in determining a state in a current measurement epoch.

17. The method of claim 15, wherein a frequency of the measurement epochs is variable and is determined at least in part based on one or more of:

a coupling pre-classification of the sensor measurements;

a probability of a coupling state of the respective device determined in a prior measurement epoch; and

a stability of a current coupling state of the respective device.

18. The method of claim 15, wherein:

the plurality of epochs include a first epoch and a second epoch;

during the first epoch;

the sensor measurements are determined to correspond to a coupling-stable pre-classification; and

a first set of feature types corresponding to the coupling-stable pre-classification are selected as the one or more selected feature types; and

during the second epoch;

the sensor measurements are determined to correspond to a coupling-transition pre-classification; and

a second set of feature types corresponding to the coupling-transition pre-classification are selected as the one or more selected feature types, wherein the first set of feature types is different from the second set of feature types.

19. The method of claim 1, further comprising, after determining the state of the respective device, adjusting operation of the respective device in accordance with the determined state of the respective device.

20. The method of claim 1, wherein:

the one or more devices include a first device and a second device; and

the method comprises;

determining a state of the first device in accordance with a classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements; and

determining a state of the second device in accordance with the classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements.

21. The method of claim 1, further comprising, prior to extracting features of the one or more feature types selected based on the pre-classification:

scheduling extraction of features of a first subset of a plurality of feature types, wherein the first subset of feature types includes the one or more feature types selected based on the pre-classification; and

scheduling extraction of features of a second subset of the plurality of feature types, wherein;

the second subset of feature types includes feature types other than the one or more feature types selected based on the pre-classification;

the extraction of features of the second subset of feature types is scheduled to occur after the extraction of features of the first subset of feature types; and

the extraction of features of the second subset of feature types is subject to cancellation.

22. The method of claim 1, further comprising:

extracting features of a first subset of a plurality of feature types, wherein the first subset includes the one or more feature types selected based on the pre-classification; and

after extracting the features of the first subset;

determining whether the features extracted from the first subset of feature types are consistent with the pre-classification;

in accordance with a determination that the features extracted from the first subset of feature types are not consistent with the pre-classification, extracting features of a second subset of feature types that includes feature types other than the one or more feature types selected based on the pre-classification; and

in accordance with a determination that the features extracted from the first subset of feature types are consistent with the pre-classification, forgoing extracting features of the second subset of feature types.

23. The method of claim 22, further comprising, prior to extracting the features of the first subset of the plurality of feature types:

scheduling extraction of features of the first subset of feature types; and

scheduling extraction of features of the second subset of feature types.

24. A processing apparatus, comprising:

one or more processors;

a set of one or more sensors;

memory; and

one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs including instructions for;

receiving sensor measurements generated by one or more sensors of one or more devices;

pre-classifying the sensor measurements as belonging to one of a plurality of pre-classifications;

selecting one or more feature types to extract from the sensor measurements based at least in part on the pre-classification of the sensor measurements, including;

in accordance with a determination that the sensor measurements correspond to a first pre-classification, selecting a first set of feature types as the one or more selected feature types; and

in accordance with a determination that the set of sensor measurements correspond to a second pre-classification, selecting a second set of feature types as the one or more selected feature types, wherein the first set of feature types is different from the second set of feature types;

extracting features of the one or more selected feature types from the sensor measurements, including;

in accordance with a determination that the sensor measurements correspond to the first pre-classification, forgoing extracting features of one or more of the second set of feature types while extracting features of the first set of feature types; and

in accordance with a determination that the set of sensor measurements correspond to the second pre-classification, forgoing extracting features of one or more of the first set of feature types while extracting features of the second set of feature types; and

determining a state of a respective device of the one or more devices in accordance with a classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements.

25. The processing apparatus of claim 24, wherein the state of the respective device comprises one or more of:

a physical property of the respective device;

a state of an environment of the respective device; and

a state of an entity associated with the respective device.

26. The processing apparatus of claim 24, wherein:

pre-classifying the sensor measurements includes identifying the sensor measurements as belonging to a respective pre-classification that is associated with multiple feature sets, including a first set of features for use in identifying states of the respective device under a first set of conditions and a second set of features for use in identifying states of the respective device under a second set of conditions different from the first set of conditions;

extracting the features includes extracting the first set of features and extracting the second set of features; and

determining the state of the respective device includes determining the state of the respective device in accordance with the first set of features and the second set of features.

27. The processing apparatus of claim 24, wherein:

pre-classifying the sensor measurements includes identifying the sensor measurements as belonging to one of a plurality of coupling pre-classifications; and

determining the state of the respective device includes determining a coupling state of the respective device in accordance with a coupling classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements.

28. The processing apparatus of claim 27, wherein the coupling state of the respective device is selected from a set consisting of:

a plurality of coupling-stable states corresponding to a coupling between the respective device and an entity associated with the respective device; and

a plurality of coupling-transition states corresponding to a transition between two of the coupling-stable states.

29. The processing apparatus of claim 24, wherein:

the first set of feature types includes features adapted for determining a state of the respective device while the respective device is stationary; and

the second set of feature types includes features adapted for determining a state of the respective device while the respective device is in motion.

30. The processing apparatus of claim 24, wherein the first set of feature types are feature types that enable classification of a coupling state of the respective device.

31. The processing apparatus of claim 24, wherein the second set of feature types are feature types that enable classification of a transition between two different coupling states of the respective device.

32. The processing apparatus of claim 24, wherein the first set of feature types and the second set of feature types are mutually exclusive.

33. The processing apparatus of claim 24, wherein the plurality of pre-classifications include:

a first pre-classification corresponding to a plurality of transition feature types associated with identifying a transition between two different states of the respective device;

a second pre-classification corresponding to a first subset of a plurality of device-state feature types associated with identifying a state of the respective device; and

a third pre-classification corresponding to a second subset of the plurality of device-state feature types, wherein the second subset of device-state feature types is different from the first subset of device-state feature types.

34. The processing apparatus of claim 24, wherein:

the one or more programs include instructions for obtaining context information indicative of a current context of the respective device; and

the sensor measurements are pre-classified based at least in part on the current context of the respective device.

35. The processing apparatus of claim 24, wherein:

pre-classifying the sensor measurements includes extracting features of one or more pre-classification feature types from the sensor measurements; and

extracting features of the pre-classification feature types from the sensor measurements is more resource efficient than extracting features of the one or more selected feature types from the sensor measurements.

36. The processing apparatus of claim 24, wherein:

the one or more selected feature types includes a respective feature type that provides an approximation of frequency bandwidth information for a respective sensor-measurement signal that corresponds to sensor measurements collected over time; and

extracting features of the one or more selected feature types from the sensor measurements includes extracting features of the respective feature type from the respective sensor-measurement signal in the time domain without converting the respective sensor-measurement signal into the frequency-domain.

37. The processing apparatus of claim 24, wherein:

determining the state of the respective device includes updating a Markov model, wherein the Markov model includes;

a plurality of model states corresponding to states of the respective device; and

a plurality of sets of transition probabilities between the plurality of model states;

in accordance with a determination that the sensor measurements correspond to a first pre-classification, updating one or more model states in the Markov model in accordance with a first set of transition probabilities; and

in accordance with a determination that the set of sensor measurements correspond to a second pre-classification, updating one or more model states in the Markov model in accordance with a second set of transition probabilities different from the first set of transition probabilities.

38. The processing apparatus of claim 24, wherein the receiving sensor measurements, pre-classifying sensor measurements, selecting one or more feature types, extracting features and determining the state of the respective device are repeated for a plurality of measurement epochs.

39. The processing apparatus of claim 38, wherein a state of the respective device determined in a prior measurement epoch is used as a factor in determining a state in a current measurement epoch.

40. The processing apparatus of claim 38, wherein a frequency of the measurement epochs is variable and is determined at least in part based on one or more of:

a coupling pre-classification of the sensor measurements;

a probability of a coupling state of the respective device determined in a prior measurement epoch; and

a stability of a current coupling state of the respective device.

41. The processing apparatus of claim 38, wherein:

the plurality of epochs include a first epoch and a second epoch;

during the first epoch;

the sensor measurements are determined to correspond to a coupling-stable pre-classification; and

a first set of feature types corresponding to the coupling-stable pre-classification are selected as the one or more selected feature types; and

during the second epoch;

the sensor measurements are determined to correspond to a coupling-transition pre-classification; and

a second set of feature types corresponding to the coupling-transition pre-classification are selected as the one or more selected feature types, wherein the first set of feature types is different from the second set of feature types.

42. The processing apparatus of claim 24, wherein the one or more programs include instructions for, after determining the state of the respective device, adjusting operation of the respective device in accordance with the determined state of the respective device.

43. The processing apparatus of claim 24, wherein:

the one or more devices include a first device and a second device; and

the one or more programs include instructions for;

determining a state of the first device in accordance with a classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements; and

determining a state of the second device in accordance with the classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements.

44. The processing apparatus of claim 24, wherein the one or more programs include instructions for, prior to extracting features of the one or more feature types selected based on the pre-classification:

scheduling extraction of features of a first subset of a plurality of feature types, wherein the first subset of feature types includes the one or more feature types selected based on the pre-classification; and

scheduling extraction of features of a second subset of the plurality of feature types, wherein;

the second subset of feature types includes feature types other than the one or more feature types selected based on the pre-classification;

the extraction of features of the second subset of feature types is scheduled to occur after the extraction of features of the first subset of feature types; and

the extraction of features of the second subset of feature types is subject to cancellation.

45. The processing apparatus of claim 24, wherein the one or more programs include instructions for:

extracting features of a first subset of a plurality of feature types, wherein the first subset includes the one or more feature types selected based on the pre-classification; and

after extracting the features of the first subset;

determining whether the features extracted from the first subset of feature types are consistent with the pre-classification;

in accordance with a determination that the features extracted from the first subset of feature types are not consistent with the pre-classification, extracting features of a second subset of feature types that includes feature types other than the one or more feature types selected based on the pre-classification; and

in accordance with a determination that the features extracted from the first subset of feature types are consistent with the pre-classification, forgoing extracting features of the second subset of feature types.

46. The processing apparatus of claim 45, wherein the one or more programs include instructions for, prior to extracting the features of the first subset of the plurality of feature types:

scheduling extraction of features of the first subset of feature types; and

scheduling extraction of features of the second subset of feature types.

47. A non-transitory computer readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by a processing apparatus with one or more processors, cause the processing apparatus to:

receive sensor measurements generated by one or more sensors of one or more devices;

pre-classify the sensor measurements as belonging to one of a plurality of pre-classifications;

select one or more feature types to extract from the sensor measurements based at least in part on the pre-classification of the sensor measurements, wherein the selecting includes;

in accordance with a determination that the sensor measurements correspond to a first pre-classification, selecting a first set of feature types as the one or more selected feature types; and

in accordance with a determination that the set of sensor measurements correspond to a second pre-classification, selecting a second set of feature types as the one or more selected feature types, wherein the first set of feature types is different from the second set of feature types;

extract features of the one or more selected feature types from the sensor measurements, wherein the extracting includes;

in accordance with a determination that the sensor measurements correspond to the first pre-classification, forgoing extracting features of one or more of the second set of feature types while extracting features of the first set of feature types; and

in accordance with a determination that the set of sensor measurements correspond to the second pre-classification, forgoing extracting features of one or more of the first set of feature types while extracting features of the second set of feature types; and

determine a state of a respective device of the one or more devices in accordance with a classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements.

48. The non-transitory computer readable storage medium of claim 47, wherein the state of the respective device comprises one or more of:

a physical property of the respective device;

a state of an environment of the respective device; and

a state of an entity associated with the respective device.

49. The non-transitory computer readable storage medium of claim 47, wherein:

pre-classifying the sensor measurements includes identifying the sensor measurements as belonging to a respective pre-classification that is associated with multiple feature sets, including a first set of features for use in identifying states of the respective device under a first set of conditions and a second set of features for use in identifying states of the respective device under a second set of conditions different from the first set of conditions;

extracting the features includes extracting the first set of features and extracting the second set of features; and

determining the state of the respective device includes determining the state of the respective device in accordance with the first set of features and the second set of features.

50. The non-transitory computer readable storage medium of claim 47, wherein:

pre-classifying the sensor measurements includes identifying the sensor measurements as belonging to one of a plurality of coupling pre-classifications; and

determining the state of the respective device includes determining a coupling state of the respective device in accordance with a coupling classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements.

51. The non-transitory computer readable storage medium of claim 50, wherein the coupling state of the respective device is selected from a set consisting of:

a plurality of coupling-stable states corresponding to a coupling between the respective device and an entity associated with the respective device; and

a plurality of coupling-transition states corresponding to a transition between two of the coupling-stable states.

52. The non-transitory computer readable storage medium of claim 47, wherein:

the first set of feature types includes features adapted for determining a state of the respective device while the respective device is stationary; and

the second set of feature types includes features adapted for determining a state of the respective device while the respective device is in motion.

53. The non-transitory computer readable storage medium of claim 47, wherein the first set of feature types are feature types that enable classification of a coupling state of the respective device.

54. The non-transitory computer readable storage medium of claim 47, wherein the second set of feature types are feature types that enable classification of a transition between two different coupling states of the respective device.

55. The non-transitory computer readable storage medium of claim 47, wherein the first set of feature types and the second set of feature types are mutually exclusive.

56. The non-transitory computer readable storage medium of claim 47, wherein the plurality of pre-classifications include:

a first pre-classification corresponding to a plurality of transition feature types associated with identifying a transition between two different states of the respective device;

a second pre-classification corresponding to a first subset of a plurality of device-state feature types associated with identifying a state of the respective device; and

a third pre-classification corresponding to a second subset of the plurality of device-state feature types, wherein the second subset of device-state feature types is different from the first subset of device-state feature types.

57. The non-transitory computer readable storage medium of claim 47, wherein:

the one or more programs include instructions which, when executed, cause the processing apparatus to obtain context information indicative of a current context of the respective device; and

the sensor measurements are pre-classified based at least in part on the current context of the respective device.

58. The non-transitory computer readable storage medium of claim 47, wherein:

pre-classifying the sensor measurements includes extracting features of one or more pre-classification feature types from the sensor measurements; and

extracting features of the pre-classification feature types from the sensor measurements is more resource efficient than extracting features of the one or more selected feature types from the sensor measurements.

59. The non-transitory computer readable storage medium of claim 47, wherein:

the one or more selected feature types includes a respective feature type that provides an approximation of frequency bandwidth information for a respective sensor-measurement signal that corresponds to sensor measurements collected over time; and

extracting features of the one or more selected feature types from the sensor measurements includes extracting features of the respective feature type from the respective sensor-measurement signal in the time domain without converting the respective sensor-measurement signal into the frequency-domain.

60. The non-transitory computer readable storage medium of claim 47, wherein:

determining the state of the respective device includes updating a Markov model, wherein the Markov model includes;

a plurality of model states corresponding to states of the respective device; and

a plurality of sets of transition probabilities between the plurality of model states;

in accordance with a determination that the sensor measurements correspond to a first pre-classification, updating one or more model states in the Markov model in accordance with a first set of transition probabilities; and

in accordance with a determination that the set of sensor measurements correspond to a second pre-classification, updating one or more model states in the Markov model in accordance with a second set of transition probabilities different from the first set of transition probabilities.

61. The non-transitory computer readable storage medium of claim 47, wherein the receiving sensor measurements, pre-classifying sensor measurements, selecting one or more feature types, extracting features and determining the state of the respective device are repeated for a plurality of measurement epochs.

62. The non-transitory computer readable storage medium of claim 61, wherein a state of the respective device determined in a prior measurement epoch is used as a factor in determining a state in a current measurement epoch.

63. The non-transitory computer readable storage medium of claim 61, wherein a frequency of the measurement epochs is variable and is determined at least in part based on one or more of:

a coupling pre-classification of the sensor measurements;

a probability of a coupling state of the respective device determined in a prior measurement epoch; and

a stability of a current coupling state of the respective device.

64. The non-transitory computer readable storage medium of claim 61, wherein:

the plurality of epochs include a first epoch and a second epoch;

during the first epoch;

the sensor measurements are determined to correspond to a coupling-stable pre-classification; and

a first set of feature types corresponding to the coupling-stable pre-classification are selected as the one or more selected feature types; and

during the second epoch;

the sensor measurements are determined to correspond to a coupling-transition pre-classification; and

a second set of feature types corresponding to the coupling-transition pre-classification are selected as the one or more selected feature types, wherein the first set of feature types is different from the second set of feature types.

65. The non-transitory computer readable storage medium of claim 47, wherein the one or more programs include instructions which, when executed, cause the processing apparatus to, after determining the state of the respective device, adjust operation of the respective device in accordance with the determined state of the respective device.

66. The non-transitory computer readable storage medium of claim 47, wherein:

the one or more devices include a first device and a second device; and

the one or more programs include instructions which, when executed, cause the processing apparatus to;

determine a state of the first device in accordance with a classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements; and

determine a state of the second device in accordance with the classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements.

67. The non-transitory computer readable storage medium of claim 47, wherein the one or more programs include instructions which, when executed, cause the processing apparatus to, prior to extracting features of the one or more feature types selected based on the pre-classification:

schedule extraction of features of a first subset of a plurality of feature types, wherein the first subset of feature types includes the one or more feature types selected based on the pre-classification; and

schedule extraction of features of a second subset of the plurality of feature types, wherein;

the second subset of feature types includes feature types other than the one or more feature types selected based on the pre-classification;

the extraction of features of the second subset of feature types is scheduled to occur after the extraction of features of the first subset of feature types; and

the extraction of features of the second subset of feature types is subject to cancellation.

68. The non-transitory computer readable storage medium of claim 47, wherein the one or more programs include instructions which, when executed, cause the processing apparatus to:

extract features of a first subset of a plurality of feature types, wherein the first subset includes the one or more feature types selected based on the pre-classification; and

after extracting the features of the first subset;

determine whether the features extracted from the first subset of feature types are consistent with the pre-classification;

in accordance with a determination that the features extracted from the first subset of feature types are not consistent with the pre-classification, extract features of a second subset of feature types that includes feature types other than the one or more feature types selected based on the pre-classification; and

in accordance with a determination that the features extracted from the first subset of feature types are consistent with the pre-classification, forgo extracting features of the second subset of feature types.

69. The non-transitory computer readable storage medium of claim 47, wherein the one or more programs include instructions which, when executed, cause the processing apparatus to, prior to extracting the features of the first subset of the plurality of feature types:

schedule extraction of features of the first subset of feature types; and

schedule extraction of features of the second subset of feature types.

70. A method comprising:

at a processing apparatus having one or more processors and memory storing one or more programs that, when executed by the one or more processors, cause the respective processing apparatus to perform the method;

receiving sensor measurements generated by one or more sensors of one or more devices;

pre-classifying the sensor measurements as belonging to one of a plurality of pre-classifications, wherein pre-classifying the sensor measurements includes identifying the sensor measurements as belonging to one of a plurality of coupling pre-classifications;

selecting one or more feature types to extract from the sensor measurements based at least in part on the pre-classification of the sensor measurements;

extracting features of the one or more selected feature types from the sensor measurements; and

determining a state of a respective device of the one or more devices in accordance with a classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements, wherein determining the state of the respective device includes determining a coupling state of the respective device in accordance with a coupling classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements, wherein the coupling state of the respective device is selected from a set consisting of;

a plurality of coupling-stable states corresponding to a coupling between the respective device and an entity associated with the respective device; and

a plurality of coupling-transition states corresponding to a transition between two of the coupling-stable states.

71. The method of claim 70, wherein selecting the one or more feature types includes:

in accordance with a determination that the sensor measurements correspond to a first pre-classification, selecting a first set of feature types as the one or more selected feature types; and

in accordance with a determination that the set of sensor measurements correspond to a second pre-classification, selecting a second set of feature types as the one or more selected feature types, wherein the first set of feature types is different from the second set of feature types.

72. The method of claim 71, wherein the first set of feature types are feature types that enable classification of a coupling state of the respective device.

73. The method of claim 72, wherein the second set of feature types are feature types that enable classification of a transition between two different coupling states of the respective device.

74. The method of claim 70, wherein the plurality of pre-classifications include:

a first pre-classification corresponding to a plurality of transition feature types associated with identifying a transition between two different states of the respective device;

a second pre-classification corresponding to a first subset of a plurality of device-state feature types associated with identifying a state of the respective device; and

a third pre-classification corresponding to a second subset of the plurality of device-state feature types, wherein the second subset of device-state feature types is different from the first subset of device-state feature types.

75. A processing apparatus, comprising:

one or more processors;

a set of one or more sensors;

memory; and

one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs including instructions for;

receiving sensor measurements generated by one or more sensors of one or more devices;

pre-classifying the sensor measurements as belonging to one of a plurality of pre-classifications, wherein pre-classifying the sensor measurements includes identifying the sensor measurements as belonging to one of a plurality of coupling pre-classifications;

selecting one or more feature types to extract from the sensor measurements based at least in part on the pre-classification of the sensor measurements;

extracting features of the one or more selected feature types from the sensor measurements; and

determining a state of a respective device of the one or more devices in accordance with a classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements, wherein determining the state of the respective device includes determining a coupling state of the respective device in accordance with a coupling classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements, wherein the coupling state of the respective device is selected from a set consisting of;

a plurality of coupling-stable states corresponding to a coupling between the respective device and an entity associated with the respective device; and

a plurality of coupling-transition states corresponding to a transition between two of the coupling-stable states.

76. The processing apparatus of claim 75, wherein selecting the one or more feature types includes:

in accordance with a determination that the sensor measurements correspond to a first pre-classification, selecting a first set of feature types as the one or more selected feature types; and

in accordance with a determination that the set of sensor measurements correspond to a second pre-classification, selecting a second set of feature types as the one or more selected feature types, wherein the first set of feature types is different from the second set of feature types.

77. The processing apparatus of claim 76, wherein the first set of feature types are feature types that enable classification of a coupling state of the respective device.

78. The processing apparatus of claim 77, wherein the second set of feature types are feature types that enable classification of a transition between two different coupling states of the respective device.

79. The processing apparatus of claim 77, wherein the plurality of pre-classifications include:

a first pre-classification corresponding to a plurality of transition feature types associated with identifying a transition between two different states of the respective device;

a second pre-classification corresponding to a first subset of a plurality of device-state feature types associated with identifying a state of the respective device; and

a third pre-classification corresponding to a second subset of the plurality of device-state feature types, wherein the second subset of device-state feature types is different from the first subset of device-state feature types.

80. A non-transitory computer readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by a processing apparatus with one or more processors, cause the processing apparatus to:

receiving sensor measurements generated by one or more sensors of one or more devices;

pre-classifying the sensor measurements as belonging to one of a plurality of pre-classifications, wherein pre-classifying the sensor measurements includes identifying the sensor measurements as belonging to one of a plurality of coupling pre-classifications;

selecting one or more feature types to extract from the sensor measurements based at least in part on the pre-classification of the sensor measurements;

extracting features of the one or more selected feature types from the sensor measurements; and

determining a state of a respective device of the one or more devices in accordance with a classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements, wherein determining the state of the respective device includes determining a coupling state of the respective device in accordance with a coupling classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements, wherein the coupling state of the respective device is selected from a set consisting of;

a plurality of coupling-stable states corresponding to a coupling between the respective device and an entity associated with the respective device; and

a plurality of coupling-transition states corresponding to a transition between two of the coupling-stable states.

81. The non-transitory computer readable storage medium of claim 80, wherein selecting the one or more feature types includes:

in accordance with a determination that the sensor measurements correspond to a first pre-classification, selecting a first set of feature types as the one or more selected feature types; and

in accordance with a determination that the set of sensor measurements correspond to a second pre-classification, selecting a second set of feature types as the one or more selected feature types, wherein the first set of feature types is different from the second set of feature types.

82. The non-transitory computer readable storage medium of claim 81, wherein the first set of feature types are feature types that enable classification of a coupling state of the respective device.

83. The non-transitory computer readable storage medium of claim 82, wherein the second set of feature types are feature types that enable classification of a transition between two different coupling states of the respective device.

84. The non-transitory computer readable storage medium of claim 82, wherein the plurality of pre-classifications include:

a first pre-classification corresponding to a plurality of transition feature types associated with identifying a transition between two different states of the respective device;

a second pre-classification corresponding to a first subset of a plurality of device-state feature types associated with identifying a state of the respective device; and

a third pre-classification corresponding to a second subset of the plurality of device-state feature types, wherein the second subset of device-state feature types is different from the first subset of device-state feature types.

85. A method comprising:

at a processing apparatus having one or more processors and memory storing one or more programs that, when executed by the one or more processors, cause the respective processing apparatus to perform the method;

receiving sensor measurements generated by one or more sensors of one or more devices;

pre-classifying the sensor measurements as belonging to one of a plurality of pre-classifications, wherein pre-classifying the sensor measurements includes extracting features of one or more pre-classification feature types from the sensor measurements;

selecting one or more feature types to extract from the sensor measurements based at least in part on the pre-classification of the sensor measurements;

extracting features of the one or more selected feature types from the sensor measurements, wherein extracting features of the pre-classification feature types from the sensor measurements is more resource efficient than extracting features of the one or more selected feature types from the sensor measurements; and

determining a state of a respective device of the one or more devices in accordance with a classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements.

86. The method of claim 85, wherein:

pre-classifying the sensor measurements includes identifying the sensor measurements as belonging to a respective pre-classification that is associated with multiple feature sets, including a first set of features for use in identifying states of the respective device under a first set of conditions and a second set of features for use in identifying states of the respective device under a second set of conditions different from the first set of conditions;

extracting the features includes extracting the first set of features and extracting the second set of features; and

determining the state of the respective device includes determining the state of the respective device in accordance with the first set of features and the second set of features.

87. The method of claim 85, wherein:

pre-classifying the sensor measurements includes identifying the sensor measurements as belonging to one of a plurality of coupling pre-classifications; and

determining the state of the respective device includes determining a coupling state of the respective device in accordance with a coupling classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements.

88. The method of claim 85, further comprising, prior to extracting features of the one or more feature types selected based on the pre-classification:

scheduling extraction of features of a first subset of a plurality of feature types, wherein the first subset of feature types includes the one or more feature types selected based on the pre-classification; and

scheduling extraction of features of a second subset of the plurality of feature types, wherein;

the second subset of feature types includes feature types other than the one or more feature types selected based on the pre-classification;

the extraction of features of the second subset of feature types is scheduled to occur after the extraction of features of the first subset of feature types; and

the extraction of features of the second subset of feature types is subject to cancellation.

89. The method of claim 85, further comprising:

extracting features of a first subset of a plurality of feature types, wherein the first subset includes the one or more feature types selected based on the pre-classification; and

after extracting the features of the first subset;

determining whether the features extracted from the first subset of feature types are consistent with the pre-classification;

in accordance with a determination that the features extracted from the first subset of feature types are not consistent with the pre-classification, extracting features of a second subset of feature types that includes feature types other than the one or more feature types selected based on the pre-classification; and

in accordance with a determination that the features extracted from the first subset of feature types are consistent with the pre-classification, forgoing extracting features of the second subset of feature types.

90. A processing apparatus, comprising:

one or more processors;

a set of one or more sensors;

memory; and

one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs including instructions for;

receiving sensor measurements generated by one or more sensors of one or more devices;

pre-classifying the sensor measurements as belonging to one of a plurality of pre-classifications, wherein pre-classifying the sensor measurements includes extracting features of one or more pre-classification feature types from the sensor measurements;

selecting one or more feature types to extract from the sensor measurements based at least in part on the pre-classification of the sensor measurements;

extracting features of the one or more selected feature types from the sensor measurements, wherein extracting features of the pre-classification feature types from the sensor measurements is more resource efficient than extracting features of the one or more selected feature types from the sensor measurements; and

determining a state of a respective device of the one or more devices in accordance with a classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements.

91. The processing apparatus of claim 90, wherein:

pre-classifying the sensor measurements includes identifying the sensor measurements as belonging to a respective pre-classification that is associated with multiple feature sets, including a first set of features for use in identifying states of the respective device under a first set of conditions and a second set of features for use in identifying states of the respective device under a second set of conditions different from the first set of conditions;

extracting the features includes extracting the first set of features and extracting the second set of features; and

determining the state of the respective device includes determining the state of the respective device in accordance with the first set of features and the second set of features.

92. The processing apparatus of claim 90, wherein:

pre-classifying the sensor measurements includes identifying the sensor measurements as belonging to one of a plurality of coupling pre-classifications; and

determining the state of the respective device includes determining a coupling state of the respective device in accordance with a coupling classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements.

93. The processing apparatus of claim 90, wherein the one or more programs include instructions for, prior to extracting features of the one or more feature types selected based on the pre-classification:

scheduling extraction of features of a first subset of a plurality of feature types, wherein the first subset of feature types includes the one or more feature types selected based on the pre-classification; and

scheduling extraction of features of a second subset of the plurality of feature types, wherein;

the second subset of feature types includes feature types other than the one or more feature types selected based on the pre-classification;

the extraction of features of the second subset of feature types is scheduled to occur after the extraction of features of the first subset of feature types; and

the extraction of features of the second subset of feature types is subject to cancellation.

94. The processing apparatus of claim 90, wherein the one or more programs include instructions for:

extracting features of a first subset of a plurality of feature types, wherein the first subset includes the one or more feature types selected based on the pre-classification; and

after extracting the features of the first subset;

determining whether the features extracted from the first subset of feature types are consistent with the pre-classification;

in accordance with a determination that the features extracted from the first subset of feature types are not consistent with the pre-classification, extracting features of a second subset of feature types that includes feature types other than the one or more feature types selected based on the pre-classification; and

in accordance with a determination that the features extracted from the first subset of feature types are consistent with the pre-classification, forgoing extracting features of the second subset of feature types.

95. A non-transitory computer readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by a processing apparatus with one or more processors, cause the processing apparatus to:

receive sensor measurements generated by one or more sensors of one or more devices;

pre-classify the sensor measurements as belonging to one of a plurality of pre-classifications, wherein pre-classifying the sensor measurements includes extracting features of one or more pre-classification feature types from the sensor measurements;

select one or more feature types to extract from the sensor measurements based at least in part on the pre-classification of the sensor measurements;

extract features of the one or more selected feature types from the sensor measurements, wherein extracting features of the pre-classification feature types from the sensor measurements is more resource efficient than extracting features of the one or more selected feature types from the sensor measurements; and

determine a state of a respective device of the one or more devices in accordance with a classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements.

96. The non-transitory computer readable storage medium of claim 95, wherein:

pre-classifying the sensor measurements includes identifying the sensor measurements as belonging to a respective pre-classification that is associated with multiple feature sets, including a first set of features for use in identifying states of the respective device under a first set of conditions and a second set of features for use in identifying states of the respective device under a second set of conditions different from the first set of conditions;

extracting the features includes extracting the first set of features and extracting the second set of features; and

determining the state of the respective device includes determining the state of the respective device in accordance with the first set of features and the second set of features.

97. The non-transitory computer readable storage medium of claim 95, wherein:

pre-classifying the sensor measurements includes identifying the sensor measurements as belonging to one of a plurality of coupling pre-classifications; and

determining the state of the respective device includes determining a coupling state of the respective device in accordance with a coupling classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements.

98. The non-transitory computer readable storage medium of claim 95, wherein the one or more programs include instructions which, when executed, cause the processing apparatus to, prior to extracting features of the one or more feature types selected based on the pre-classification:

schedule extraction of features of a first subset of a plurality of feature types, wherein the first subset of feature types includes the one or more feature types selected based on the pre-classification; and

schedule extraction of features of a second subset of the plurality of feature types, wherein;

the second subset of feature types includes feature types other than the one or more feature types selected based on the pre-classification;

the extraction of features of the second subset of feature types is scheduled to occur after the extraction of features of the first subset of feature types; and

the extraction of features of the second subset of feature types is subject to cancellation.

99. The non-transitory computer readable storage medium of claim 95, wherein the one or more programs include instructions which, when executed, cause the processing apparatus to:

extract features of a first subset of a plurality of feature types, wherein the first subset includes the one or more feature types selected based on the pre-classification; and

after extracting the features of the first subset;

determine whether the features extracted from the first subset of feature types are consistent with the pre-classification;

in accordance with a determination that the features extracted from the first subset of feature types are not consistent with the pre-classification, extract features of a second subset of feature types that includes feature types other than the one or more feature types selected based on the pre-classification; and

in accordance with a determination that the features extracted from the first subset of feature types are consistent with the pre-classification, forgo extracting features of the second subset of feature types.

100. A method comprising:

at a processing apparatus having one or more processors and memory storing one or more programs that, when executed by the one or more processors, cause the respective processing apparatus to perform the method;

receiving sensor measurements generated by one or more sensors of one or more devices;

pre-classifying the sensor measurements as belonging to one of a plurality of pre-classifications;

selecting one or more feature types to extract from the sensor measurements based at least in part on the pre-classification of the sensor measurements;

extracting features of the one or more selected feature types from the sensor measurements;

determining a state of a respective device of the one or more devices in accordance with a classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements, wherein determining the state of the respective device includes updating a Markov model, wherein;

the Markov model includes a plurality of model states corresponding to states of the respective device and a plurality of sets of transition probabilities between the plurality of model states; and

updating the Markov model includes;

in accordance with a determination that the sensor measurements correspond to a first pre-classification, updating one or more model states in the Markov model in accordance with a first set of transition probabilities; and

in accordance with a determination that the set of sensor measurements correspond to a second pre-classification, updating one or more model states in the Markov model in accordance with a second set of transition probabilities different from the first set of transition probabilities.

101. The method of claim 100, wherein the receiving sensor measurements, pre-classifying sensor measurements, selecting one or more feature types, extracting features and determining the state of the respective device are repeated for a plurality of measurement epochs.

102. The method of claim 101, wherein a state of the respective device determined in a prior measurement epoch is used as a factor in determining a state in a current measurement epoch.

103. The method of claim 101, wherein a frequency of the measurement epochs is variable and is determined at least in part based on one or more of:

a coupling pre-classification of the sensor measurements;

a probability of a coupling state of the respective device determined in a prior measurement epoch; and

a stability of a current coupling state of the respective device.

104. The method of claim 101, wherein:

the plurality of epochs include a first epoch and a second epoch;

during the first epoch;

the sensor measurements are determined to correspond to a coupling-stable pre-classification; and

a first set of feature types corresponding to the coupling-stable pre-classification are selected as the one or more selected feature types; and

during the second epoch;

the sensor measurements are determined to correspond to a coupling-transition pre-classification; and

a second set of feature types corresponding to the coupling-transition pre-classification are selected as the one or more selected feature types, wherein the first set of feature types is different from the second set of feature types.

105. A processing apparatus, comprising:

one or more processors;

a set of one or more sensors;

memory; and

one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs including instructions for;

receiving sensor measurements generated by one or more sensors of one or more devices;

pre-classifying the sensor measurements as belonging to one of a plurality of pre-classifications;

selecting one or more feature types to extract from the sensor measurements based at least in part on the pre-classification of the sensor measurements;

extracting features of the one or more selected feature types from the sensor measurements;

determining a state of a respective device of the one or more devices in accordance with a classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements, wherein determining the state of the respective device includes updating a Markov model, wherein;

the Markov model includes a plurality of model states corresponding to states of the respective device and a plurality of sets of transition probabilities between the plurality of model states; and

updating the Markov model includes;

in accordance with a determination that the sensor measurements correspond to a first pre-classification, updating one or more model states in the Markov model in accordance with a first set of transition probabilities; and

in accordance with a determination that the set of sensor measurements correspond to a second pre-classification, updating one or more model states in the Markov model in accordance with a second set of transition probabilities different from the first set of transition probabilities.

106. The processing apparatus of claim 105, wherein the receiving sensor measurements, pre-classifying sensor measurements, selecting one or more feature types, extracting features and determining the state of the respective device are repeated for a plurality of measurement epochs.

107. The processing apparatus of claim 106, wherein a state of the respective device determined in a prior measurement epoch is used as a factor in determining a state in a current measurement epoch.

108. The processing apparatus of claim 106, wherein a frequency of the measurement epochs is variable and is determined at least in part based on one or more of:

a coupling pre-classification of the sensor measurements;

a probability of a coupling state of the respective device determined in a prior measurement epoch; and

a stability of a current coupling state of the respective device.

109. The processing apparatus of claim 106, wherein:

the plurality of epochs include a first epoch and a second epoch;

during the first epoch;

the sensor measurements are determined to correspond to a coupling-stable pre-classification; and

a first set of feature types corresponding to the coupling-stable pre-classification are selected as the one or more selected feature types; and

during the second epoch;

the sensor measurements are determined to correspond to a coupling-transition pre-classification; and

a second set of feature types corresponding to the coupling-transition pre-classification are selected as the one or more selected feature types, wherein the first set of feature types is different from the second set of feature types.

110. A non-transitory computer readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by a processing apparatus with one or more processors, cause the processing apparatus to:

receive sensor measurements generated by one or more sensors of one or more devices;

pre-classify the sensor measurements as belonging to one of a plurality of pre-classifications;

select one or more feature types to extract from the sensor measurements based at least in part on the pre-classification of the sensor measurements;

extract features of the one or more selected feature types from the sensor measurements;

determine a state of a respective device of the one or more devices in accordance with a classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements, wherein determining the state of the respective device includes updating a Markov model, wherein;

the Markov model includes a plurality of model states corresponding to states of the respective device and a plurality of sets of transition probabilities between the plurality of model states; and

updating the Markov model includes;

in accordance with a determination that the sensor measurements correspond to a first pre-classification, updating one or more model states in the Markov model in accordance with a first set of transition probabilities; and

in accordance with a determination that the set of sensor measurements correspond to a second pre-classification, updating one or more model states in the Markov model in accordance with a second set of transition probabilities different from the first set of transition probabilities.

111. The non-transitory computer readable storage medium of claim 110, wherein the receiving sensor measurements, pre-classifying sensor measurements, selecting one or more feature types, extracting features and determining the state of the respective device are repeated for a plurality of measurement epochs.

112. The non-transitory computer readable storage medium of claim 111, wherein a state of the respective device determined in a prior measurement epoch is used as a factor in determining a state in a current measurement epoch.

113. The non-transitory computer readable storage medium of claim 111, wherein a frequency of the measurement epochs is variable and is determined at least in part based on one or more of:

a coupling pre-classification of the sensor measurements;

a probability of a coupling state of the respective device determined in a prior measurement epoch; and

a stability of a current coupling state of the respective device.

114. The non-transitory computer readable storage medium of claim 111, wherein:

the plurality of epochs include a first epoch and a second epoch;

during the first epoch;

the sensor measurements are determined to correspond to a coupling-stable pre-classification; and

a first set of feature types corresponding to the coupling-stable pre-classification are selected as the one or more selected feature types; and

during the second epoch;

the sensor measurements are determined to correspond to a coupling-transition pre-classification; and

a second set of feature types corresponding to the coupling-transition pre-classification are selected as the one or more selected feature types, wherein the first set of feature types is different from the second set of feature types.

115. A method comprising:

at a processing apparatus having one or more processors and memory storing one or more programs that, when executed by the one or more processors, cause the respective processing apparatus to perform the method;

receiving sensor measurements generated by one or more sensors of one or more devices;

pre-classifying the sensor measurements as belonging to one of a plurality of pre-classifications;

selecting one or more feature types to extract from the sensor measurements based at least in part on the pre-classification of the sensor measurements;

extracting features of the one or more selected feature types from the sensor measurements;

determining a state of a respective device of the one or more devices in accordance with a classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements; and

wherein;

the receiving sensor measurements, pre-classifying sensor measurements, selecting one or more feature types, extracting features and determining the state of the respective device are repeated for a plurality of measurement epochs that include a first epoch and a second epoch;

during the first epoch;

the sensor measurements are determined to correspond to a coupling-stable pre-classification; and

a first set of feature types corresponding to the coupling-stable pre-classification are selected as the one or more selected feature types; and

during the second epoch;

the sensor measurements are determined to correspond to a coupling-transition pre-classification; and

a second set of feature types corresponding to the coupling-transition pre-classification are selected as the one or more selected feature types, wherein the first set of feature types is different from the second set of feature types.

116. The method of claim 115, wherein:

the first set of feature types includes features adapted for determining a state of the respective device while the respective device is stationary; and

the second set of feature types includes features adapted for determining a state of the respective device while the respective device is in motion.

117. The method of claim 115, wherein the first set of feature types are feature types that enable classification of a coupling state of the respective device.

118. The method of claim 117, wherein the second set of feature types are feature types that enable classification of a transition between two different coupling states of the respective device.

119. The method of claim 115, wherein a frequency of the measurement epochs is variable and is determined at least in part based on one or more of:

a coupling pre-classification of the sensor measurements;

a probability of a coupling state of the respective device determined in a prior measurement epoch; and

a stability of a current coupling state of the respective device.

120. A processing apparatus, comprising:

one or more processors;

a set of one or more sensors;

memory; and

one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs including instructions for;

receiving sensor measurements generated by one or more sensors of one or more devices;

pre-classifying the sensor measurements as belonging to one of a plurality of pre-classifications;

selecting one or more feature types to extract from the sensor measurements based at least in part on the pre-classification of the sensor measurements;

extracting features of the one or more selected feature types from the sensor measurements;

determining a state of a respective device of the one or more devices in accordance with a classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements; and

wherein;

the receiving sensor measurements, pre-classifying sensor measurements, selecting one or more feature types, extracting features and determining the state of the respective device are repeated for a plurality of measurement epochs that include a first epoch and a second epoch;

during the first epoch;

the sensor measurements are determined to correspond to a coupling-stable pre-classification; and

a first set of feature types corresponding to the coupling-stable pre-classification are selected as the one or more selected feature types; and

during the second epoch;

the sensor measurements are determined to correspond to a coupling-transition pre-classification; and

a second set of feature types corresponding to the coupling-transition pre-classification are selected as the one or more selected feature types, wherein the first set of feature types is different from the second set of feature types.

121. The processing apparatus of claim 120, wherein:

the first set of feature types includes features adapted for determining a state of the respective device while the respective device is stationary; and

the second set of feature types includes features adapted for determining a state of the respective device while the respective device is in motion.

122. The processing apparatus of claim 120, wherein the first set of feature types are feature types that enable classification of a coupling state of the respective device.

123. The processing apparatus of claim 122, wherein the second set of feature types are feature types that enable classification of a transition between two different coupling states of the respective device.

124. The processing apparatus of claim 120, wherein a frequency of the measurement epochs is variable and is determined at least in part based on one or more of:

a coupling pre-classification of the sensor measurements;

a probability of a coupling state of the respective device determined in a prior measurement epoch; and

a stability of a current coupling state of the respective device.

125. A non-transitory computer readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by a processing apparatus with one or more processors, cause the processing apparatus to:

receive sensor measurements generated by one or more sensors of one or more devices;

pre-classify the sensor measurements as belonging to one of a plurality of pre-classifications;

select one or more feature types to extract from the sensor measurements based at least in part on the pre-classification of the sensor measurements;

extract features of the one or more selected feature types from the sensor measurements;

determine a state of a respective device of the one or more devices in accordance with a classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements; and

wherein;

the receiving sensor measurements, pre-classifying sensor measurements, selecting one or more feature types, extracting features and determining the state of the respective device are repeated for a plurality of measurement epochs that include a first epoch and a second epoch;

during the first epoch;

the sensor measurements are determined to correspond to a coupling-stable pre-classification; and

a first set of feature types corresponding to the coupling-stable pre-classification are selected as the one or more selected feature types; and

during the second epoch;

the sensor measurements are determined to correspond to a coupling-transition pre-classification; and

a second set of feature types corresponding to the coupling-transition pre-classification are selected as the one or more selected feature types, wherein the first set of feature types is different from the second set of feature types.

126. The non-transitory computer readable storage medium of claim 125, wherein:

the first set of feature types includes features adapted for determining a state of the respective device while the respective device is stationary; and

the second set of feature types includes features adapted for determining a state of the respective device while the respective device is in motion.

127. The non-transitory computer readable storage medium of claim 125, wherein the first set of feature types are feature types that enable classification of a coupling state of the respective device.

128. The non-transitory computer readable storage medium of claim 127, wherein the second set of feature types are feature types that enable classification of a transition between two different coupling states of the respective device.

129. The non-transitory computer readable storage medium of claim 125, wherein a frequency of the measurement epochs is variable and is determined at least in part based on one or more of:

a coupling pre-classification of the sensor measurements;

a probability of a coupling state of the respective device determined in a prior measurement epoch; and

a stability of a current coupling state of the respective device.

130. A method comprising:

at a processing apparatus having one or more processors and memory storing one or more programs that, when executed by the one or more processors, cause the respective processing apparatus to perform the method;

receiving sensor measurements generated by one or more sensors of one or more devices;

pre-classifying the sensor measurements as belonging to one of a plurality of pre-classifications;

selecting one or more feature types to extract from the sensor measurements based at least in part on the pre-classification of the sensor measurements;

scheduling extraction of features of a first subset of a plurality of feature types, wherein the first subset of feature types includes the one or more feature types selected based on the pre-classification; and

scheduling extraction of features of a second subset of the plurality of feature types, wherein;

the second subset of feature types includes feature types other than the one or more feature types selected based on the pre-classification;

the extraction of features of the second subset of feature types is scheduled to occur after the extraction of features of the first subset of feature types; and

the extraction of features of the second subset of feature types is subject to cancellation;

extracting features of the one or more selected feature types from the sensor measurements; and

determining a state of a respective device of the one or more devices in accordance with a classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements.

131. The method of claim 130, wherein:

the first subset of feature types includes features adapted for determining a state of the respective device while the respective device is stationary; and

the second subset of feature types includes features adapted for determining a state of the respective device while the respective device is in motion.

132. The method of claim 130, wherein the first subset of feature types are feature types that enable classification of a coupling state of the respective device.

133. The method of claim 132, wherein the second subset of feature types are feature types that enable classification of a transition between two different coupling states of the respective device.

134. The method of claim 130, wherein the first subset of feature types and the second subset of feature types are mutually exclusive.

135. A processing apparatus, comprising:

one or more processors;

a set of one or more sensors;

memory; and

one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs including instructions for;

receiving sensor measurements generated by one or more sensors of one or more devices;

pre-classifying the sensor measurements as belonging to one of a plurality of pre-classifications;

selecting one or more feature types to extract from the sensor measurements based at least in part on the pre-classification of the sensor measurements;

scheduling extraction of features of a first subset of a plurality of feature types, wherein the first subset of feature types includes the one or more feature types selected based on the pre-classification; and

scheduling extraction of features of a second subset of the plurality of feature types, wherein;

the second subset of feature types includes feature types other than the one or more feature types selected based on the pre-classification;

the extraction of features of the second subset of feature types is scheduled to occur after the extraction of features of the first subset of feature types; and

the extraction of features of the second subset of feature types is subject to cancellation;

extracting features of the one or more selected feature types from the sensor measurements; and

determining a state of a respective device of the one or more devices in accordance with a classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements.

136. The processing apparatus of claim 135, wherein:

the first subset of feature types includes features adapted for determining a state of the respective device while the respective device is stationary; and

the second subset of feature types includes features adapted for determining a state of the respective device while the respective device is in motion.

137. The processing apparatus of claim 135, wherein the first subset of feature types are feature types that enable classification of a coupling state of the respective device.

138. The processing apparatus of claim 137, wherein the second subset of feature types are feature types that enable classification of a transition between two different coupling states of the respective device.

139. The processing apparatus of claim 135, wherein the first subset of feature types and the second subset of feature types are mutually exclusive.

140. A non-transitory computer readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by a processing apparatus with one or more processors, cause the processing apparatus to:

receive sensor measurements generated by one or more sensors of one or more devices;

pre-classify the sensor measurements as belonging to one of a plurality of pre-classifications;

select one or more feature types to extract from the sensor measurements based at least in part on the pre-classification of the sensor measurements;

schedule extraction of features of a first subset of a plurality of feature types, wherein the first subset of feature types includes the one or more feature types selected based on the pre-classification; and

schedule extraction of features of a second subset of the plurality of feature types, wherein;

the second subset of feature types includes feature types other than the one or more feature types selected based on the pre-classification;

the extraction of features of the second subset of feature types is scheduled to occur after the extraction of features of the first subset of feature types; and

the extraction of features of the second subset of feature types is subject to cancellation;

extract features of the one or more selected feature types from the sensor measurements; and

determine a state of a respective device of the one or more devices in accordance with a classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements.

141. The non-transitory computer readable storage medium of claim 140, wherein:

the first subset of feature types includes features adapted for determining a state of the respective device while the respective device is stationary; and

the second subset of feature types includes features adapted for determining a state of the respective device while the respective device is in motion.

142. The non-transitory computer readable storage medium of claim 140, wherein the first subset of feature types are feature types that enable classification of a coupling state of the respective device.

143. The non-transitory computer readable storage medium of claim 142, wherein the second subset of feature types are feature types that enable classification of a transition between two different coupling states of the respective device.

144. The non-transitory computer readable storage medium of claim 140, wherein the first subset of feature types and the second subset of feature types are mutually exclusive.

145. A method comprising:

at a processing apparatus having one or more processors and memory storing one or more programs that, when executed by the one or more processors, cause the respective processing apparatus to perform the method;

receiving sensor measurements generated by one or more sensors of one or more devices;

pre-classifying the sensor measurements as belonging to one of a plurality of pre-classifications;

selecting one or more feature types to extract from the sensor measurements based at least in part on the pre-classification of the sensor measurements;

extracting features of the one or more selected feature types from the sensor measurements, wherein extracting the features of the one or more selected feature types includes extracting features of a first subset of a plurality of feature types, wherein the first subset includes the one or more feature types selected based on the pre-classification; and

after extracting the features of the first subset;

determining whether the features extracted from the first subset of feature types are consistent with the pre-classification;

in accordance with a determination that the features extracted from the first subset of feature types are not consistent with the pre-classification, extracting features of a second subset of feature types that includes feature types other than the one or more feature types selected based on the pre-classification; and

in accordance with a determination that the features extracted from the first subset of feature types are consistent with the pre-classification, forgoing extracting features of the second subset of feature types; and

determining a state of a respective device of the one or more devices in accordance with a classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements.

146. The method of claim 145, wherein:

the first subset of feature types includes features adapted for determining a state of the respective device while the respective device is stationary; and

the second subset of feature types includes features adapted for determining a state of the respective device while the respective device is in motion.

147. The method of claim 145, wherein the first subset of feature types are feature types that enable classification of a coupling state of the respective device.

148. The method of claim 147, wherein the second subset of feature types are feature types that enable classification of a transition between two different coupling states of the respective device.

149. The method of claim 145, further comprising, prior to extracting the features of the first subset of the plurality of feature types:

scheduling extraction of features of the first subset of feature types; and

scheduling extraction of features of the second subset of feature types.

150. A processing apparatus, comprising:

one or more processors;

a set of one or more sensors;

memory; and

one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs including instructions for;

receiving sensor measurements generated by one or more sensors of one or more devices;

pre-classifying the sensor measurements as belonging to one of a plurality of pre-classifications;

selecting one or more feature types to extract from the sensor measurements based at least in part on the pre-classification of the sensor measurements;

extracting features of the one or more selected feature types from the sensor measurements, wherein extracting the features of the one or more selected feature types includes extracting features of a first subset of a plurality of feature types, wherein the first subset includes the one or more feature types selected based on the pre-classification; and

after extracting the features of the first subset;

determining whether the features extracted from the first subset of feature types are consistent with the pre-classification;

in accordance with a determination that the features extracted from the first subset of feature types are not consistent with the pre-classification, extracting features of a second subset of feature types that includes feature types other than the one or more feature types selected based on the pre-classification; and

in accordance with a determination that the features extracted from the first subset of feature types are consistent with the pre-classification, forgoing extracting features of the second subset of feature types; and

determining a state of a respective device of the one or more devices in accordance with a classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements.

151. The processing apparatus of claim 150, wherein:

the first subset of feature types includes features adapted for determining a state of the respective device while the respective device is stationary; and

the second subset of feature types includes features adapted for determining a state of the respective device while the respective device is in motion.

152. The processing apparatus of claim 150, wherein the first subset of feature types are feature types that enable classification of a coupling state of the respective device.

153. The processing apparatus of claim 152, wherein the second subset of feature types are feature types that enable classification of a transition between two different coupling states of the respective device.

154. The processing apparatus of claim 150, wherein the one or more programs include instructions for, prior to extracting the features of the first subset of the plurality of feature types:

scheduling extraction of features of the first subset of feature types; and

scheduling extraction of features of the second subset of feature types.

155. A non-transitory computer readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by a processing apparatus with one or more processors, cause the processing apparatus to:

receive sensor measurements generated by one or more sensors of one or more devices;

pre-classify the sensor measurements as belonging to one of a plurality of pre-classifications;

select one or more feature types to extract from the sensor measurements based at least in part on the pre-classification of the sensor measurements;

extract features of the one or more selected feature types from the sensor measurements, wherein extracting the features of the one or more selected feature types includes extracting features of a first subset of a plurality of feature types, wherein the first subset includes the one or more feature types selected based on the pre-classification; and

after extracting the features of the first subset;

determine whether the features extracted from the first subset of feature types are consistent with the pre-classification;

in accordance with a determination that the features extracted from the first subset of feature types are not consistent with the pre-classification, extract features of a second subset of feature types that includes feature types other than the one or more feature types selected based on the pre-classification; and

in accordance with a determination that the features extracted from the first subset of feature types are consistent with the pre-classification, forgo extracting features of the second subset of feature types; and

determine a state of a respective device of the one or more devices in accordance with a classification of the sensor measurements determined based on the one or more features extracted from the sensor measurements.

156. The non-transitory computer readable storage medium of claim 155, wherein:

the first subset of feature types includes features adapted for determining a state of the respective device while the respective device is stationary; and

the second subset of feature types includes features adapted for determining a state of the respective device while the respective device is in motion.

157. The non-transitory computer readable storage medium of claim 155, wherein the first subset of feature types are feature types that enable classification of a coupling state of the respective device.

158. The non-transitory computer readable storage medium of claim 157, wherein the second subset of feature types are feature types that enable classification of a transition between two different coupling states of the respective device.

159. The non-transitory computer readable storage medium of claim 155, wherein the one or more programs include instructions which, when executed, cause the processing apparatus to, prior to extracting the features of the first subset of the plurality of feature types:

schedule extraction of features of the first subset of feature types; and

schedule extraction of features of the second subset of feature types.