System and method for learning patterns of behavior and operating a monitoring and response system based thereon

• US 20030229471A1
• Filed: 01/10/2003
• Published: 12/11/2003
• Est. Priority Date: 01/22/2002
• Status: Active Grant

• Current Assignee: Honeywell International Inc.
• Original Assignee: Honeywell International Inc.
• Inventors: Guralnik, Valerie, Haigh, Karen Z., Harp, Steven A.
• Granted Patent: US 7,552,030 B2
• Time in Patent Office: Days
• US Class Current: 702/182
• CPC Class Codes: G06F 2218/00 Aspects of pattern recognition specially adapted for signal processing
• CPC Class Codes: G06N 5/00 Computing arrangements using knowledge-based models
• CPC Class Codes: G08B 21/0423 detecting deviation from an expected pattern of behaviour or schedule

Abstract

A system and method of operating a monitoring and response system for an actor in a daily living environment that relies upon learned models of behavior for adapting system operation. The learned model of behavior preferably includes sequential patterns organized pursuant to assigned partition values that in turn are generated based upon an evaluation of accumulated data. Based upon reference to the learned model of behavior, the system can generate more appropriate response plans based upon expected or unexpected activities, more readily recognize intended activities, recognize abandoned tasks, formulate probabilities of method choice, build probabilities of action success, anticipate and respond to actor movement within the environment, optimize response plan effectiveness, and share learned models across two or more separate system installations.

Claims

1. A method of operating an automated monitoring system adapted to monitor an actor in an environment, the method comprising:

providing at least one sensor for monitoring at least one of the actor and the environment;

accumulating information signaled by the sensor over time, the sensor information being indicative of functioning of at least one of the actor, the environment, and the system; and

automatically generating a learned model of behavior based upon the accumulated sensor information.

2. The method of claim 1, wherein the environment is a daily living environment.

3. The method of claim 1, wherein the functioning indicated by the accumulated sensor information is selected from the group consisting of an activity, a preference, a status description, and performance.

4. The method of claim 3, wherein the functioning indicated by the accumulated sensor information relates to a subject selected from the group consisting of the actor, the environment, and the system.

5. The method of claim 1, wherein the learned model provides at least one property selected from the group consisting of time of day, day of week, day of month, day of year, month, year, duration, location, frequency, likelihood of activity occurring, likelihood of activity not occurring, likelihood of commencing activity, likelihood of abandoning activity, possible outcomes, and effectiveness of outcomes.

6. The method of claim 1, wherein the learned model provides at least one property selected from the group consisting of frequency of correlation between functionings, dependence of the functionings, duration between start times of the functionings, duration between end times of the functionings, duration between specific points of the functionings, and effect of a first functioning on a second functioning.

7. The method of claim 1, further comprising:

generating a plurality of learned models of behavior each relating to a different subject matter.

8. The method of claim 1, further comprising:

providing the learned model of behavior to a second monitoring system relating to at least one of a second actor and a second environment; and

operating the second system utilizing the learned model of behavior as a baseline.

9. The method of claim 1, further comprising:

creating a meta-model of behavior by merging at least two learned models of behavior; and

operating a monitoring system utilizing the meta-model of behavior as a baseline.

10. The method of claim 1, further comprising:

storing feedback information relating to at least one of the actor and the environment;

wherein generating the learned model of behavior is based upon the stored feedback information.

11. The method of claim 10, wherein the feedback information relates to a preference of the actor.

12. The method of claim 10, wherein the feedback information relates to performance of the system.

13. The method of claim 1, further comprising:

posing questions; and

recording responses to the questions;

wherein the learned model of behavior is based upon the recorded responses.

14. The method of claim 13, wherein the questions relate to a status of at least one of the actor and the environment.

15. The method of claim 13, wherein the questions relates to an activity of at least one of the actor and the environment.

16. The method of claim 1, wherein the learned model of behavior is indicative of an expected activity, the method further comprising:

determining a current activity of at least one of the actor and the environment; and

comparing the current activity with the expected activity.

17. The method of claim 1, further comprising:

adapting operation of the system based upon the learned model of behavior.

18. The method of claim 1, wherein automatically generating a learned model of behavior further includes:

condensing multiple, sequential firings of the sensor into a single data item.

19. The method of claim 1, wherein a plurality of sensors are provided, the method further comprising:

condensing firings of two or more of the plurality of sensors into a single data item.

20. The method of claim 1, further comprising:

automatically assessing a current situation of at least one of the actor and the environment based in part upon the learned model of behavior.

21. The method of claim 20, wherein the system includes an intent recognition module adapted to determine an intended activity of the actor based upon a series of sensed actions, and further wherein assessing a current situation includes:

providing the learned model of behavior to the intent recognition module; and

operating the intent recognition module to designate an intended activity of the actor based upon the sensed actions and the learned model of behavior.

22. The method of claim 20, wherein the learned model of behavior is indicative of steps normally performed by the actor in completing a particular task, and further wherein assessing a current situation includes:

determining that the actor is engaged in the particular task;

comparing steps actually performed by the actor as part of the engaged particular task with the learned model of behavior; and

determining that the actor has abandoned the particular task based upon the comparison.

23. The method of claim 1, wherein the learned model of behavior is indicative of a probability of the actor selecting a particular one of a plurality of action options.

24. The method of claim 1, further comprising:

operating the system to generate a response based upon reference to the learned model of behavior.

25. The method of claim 24, wherein the learned model of behavior is indicative of an expected activity, and further wherein operating the system includes generating a response when a current activity varies from the expected activity.

26. The method of claim 24, wherein operating the system to generate a response includes:

adapting a generated response based upon reference to the learned model of behavior.

27. The method of claim 26, wherein the learned model of behavior is indicative of an effectiveness of previously generated responses, and further wherein adapting a generated response includes:

formatting a response based upon the effectiveness of previously generated responses.

28. The method of claim 27, wherein the response includes a message component for providing a message to at least one of the actor and other humans, and further wherein the learned model of behavior is indicative of an effectiveness of previous messages, and further wherein formatting the response includes:

formulating the message component based upon the effectiveness of previous messages.

29. The method of claim 24, wherein the learned model of behavior is indicative of the actor selecting a particular one of a plurality of action options, and further wherein operating the system to generate a response includes:

generating a response based upon the probability of action selection.

30. The method of claim 24, wherein the learned model of behavior is indicative of a likelihood of the actor successfully completing an activity, and further wherein operating the system to generate a response includes:

generating a response based upon the likelihood of the actor successfully completing the activity.

31. The method of claim 24, wherein the learned model of behavior is indicative of an expected path the actor normally follows in the environment, and further wherein operating the system includes:

generating a response based upon the expected path.

32. The method of claim 31, wherein the generated response includes actuating at least one environment component along the expected path.

33. The method of claim 1, wherein automatically generating a learned model of behavior includes:

organizing the accumulated data into sequential patterns.

34. The method of claim 33, wherein a plurality of sensors are provided, and further wherein organizing the accumulated data includes:

designating discrete partitions for each of the sensors; and

categorizing information from individual ones of the sensors in accordance with the designated partitions;

wherein the sequential patterns are formulated based upon the categorized information.

35. The method of claim 34, wherein the discrete partitions are time interval values, and further wherein designating time interval values includes:

separately aggregating sensor information for each sensor; and

determining at least one time interval value for each sensor based upon the respective aggregated sensor information.

36. The method of claim 35, wherein determining at least one time interval value includes:

estimating a probability density function for aggregated sensor information.

37. The method of claim 33, wherein automatically generating a learned model of behavior further includes:

eliminating redundant sequential patterns from the learned model of behavior.

38. The method of claim 37, wherein eliminating redundant sequential patterns includes:

identifying sequential sub-patterns; and

removing all sequential sub-patterns that have identical days of occurrence as a corresponding sequential pattern.

39. A system for automatically monitoring and supporting an actor in an environment, the system comprising:

at least one sensor;

at least one effector; and

a controller for receiving information from the sensor to monitor the actor and for controlling operations of the effector based upon the monitored information, the controller further being adapted to;

accumulate information signaled by the sensor over time, automatically generate a learned model of behavior relating to one of the actor and the environment based upon the accumulated sensor information.

40. The system of claim 39, wherein the environment is a daily living environment.

41. The system of claim 39, wherein the learned model of behavior relates to a status of at least one of the actor and the environment.

42. The system of claim 39, wherein the learned model of behavior relates to an activity of at least one of the actor and the environment.

43. The system of claim 39, wherein the learned model of behavior relates to a preference of at least one of the actor and an other person in the environment.

44. The system of claim 39, wherein the learned model of behavior relates to performance of the system.

45. The system of claim 39, wherein the controller is further adapted to:

generate a plurality of learned models of behavior each relating to a different subject matter.

46. The system of claim 39, wherein the controller is further adapted to:

receive a second learned model of behavior from a second monitoring and response system; and

operate the system utilizing the second learned model of behavior as a baseline.

47. The system of claim 39, wherein the learned model of behavior is indicative of an expected activity of one of the actor and the environment, and further wherein the controller is adapted to:

determine a current activity of one of the actor and the environment; and

compare the current activity with the expected activity.

48. The system of claim 39, wherein the controller is further adapted to:

adapt operation of the system based upon the learned model of behavior.

49. The system of claim 39, wherein the controller is further adapted to:

assess a current situation of at least one of the actor and the environment based, in part, upon the learned model of behavior.

50. The system of claim 39, wherein the learned model of behavior is indicative of a likelihood of the actor successfully completing a particular activity, and further wherein the controller is adapted to:

recognize that the actor is engaged in the particular activity; and

generate a response relating to the engaged particular activity based upon the learned model of behavior.

51. The system of claim 39, wherein the learned model of behavior is indicative of effectiveness of previously generated responses, and further wherein the controller is adapted to:

generate a current response based upon the effectiveness of previously generated responses.

52. The system of claim 39, wherein the controller is further adapted to:

organize the accumulated data into sequential data patterns.

53. The system of claim 52, wherein the system includes a plurality of sensors, and wherein the controller is further adapted to:

designate discrete partitions for each of the sensors; and

categorize the information from individual ones of the sensors in accordance with the designated partitions;

wherein the sequential patterns are formulated based upon the categorized information.

54. The system of claim 53, wherein the discrete partitions are time interval values and wherein the controller is further adapted to:

aggregate sensor information for each sensor; and

determine at least one time interval value for each sensor based upon the respective aggregated sensor information.

55. The system of claim 54, wherein the controller is further adapted to:

estimate a probability density function for each of the aggregated sensor information.

56. The system of claim 52, wherein the controller is further adapted to:

eliminate redundant sequential patterns from the learned model of behavior.

57. The system of claim 56, wherein the controller is further adapted to:

identify sequential sub-patterns; and

remove all sequential sub-patterns from the learned model of behavior that have identical days of occurrence as a corresponding sequential pattern.

58. A method of building a learned model of behavior for use by an actor monitoring and support system, the method comprising:

receiving data items for at least two weeks from a plurality of information sources each relating to at least one of the actor or an environment of the actor;

separately aggregating the received data items for each information source;

determining discrete partitions for each of the information sources based upon an evaluation of the corresponding, aggregated data; and

assigning a corresponding partition value to each received data item.

59. The method of claim 58, further comprising:

organizing the received data items into sequential patterns based upon the assigned partition values.

60. The method of claim 58, wherein determining partition values for each of the information sources includes estimating a probability density function for the aggregated data.