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

• US 7,552,030 B2
• Filed: 01/10/2003
• Issued: 06/23/2009
• Est. Priority Date: 01/22/2002
• Status: Active Grant

• Current Assignee: Honeywell International Inc.
• Original Assignee: Honeywell International Inc.
• Inventors: Harp, Steven A., Guralnik, Valerie, Haigh, Karen Z.
• Primary Examiner(s): Feliciano; Eliseo Ramos
• Assistant Examiner(s): Huynh; Phuong
• Application Number:
  US10/339,941
• Publication Number:
  US 20030229471A1
• Time in Patent Office: 2,356 Days
• Field of Search: 702179-190, 702/19, 700/9, 700/10, 714/39, 703/2
• US Class Current: 702/188
• 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 said actor and said environment;

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

determining a probability density function for said information signaled by each sensor of said at least one sensor in order to create data indicative of said each sensor activity;

utilizing a plurality of filter routines to reduce a quantity of said data indicative of said each sensor activity wherein said plurality of filter routines includes a plurality of the following;

a domain heuristic filter routine, a redundant pattern filter routine, a repetitive sensor firing routine and a cluster technique filter routine;

automatically generating a learned model of behavior based upon said data indicative of said each sensor activity reduced by said plurality of filter routines;

operating said system to generate a response for assisting said actor based upon reference to said learned model of behavior; and

automatically acting upon said response to assist said actor, characterized by altering a current mode of operation of a user interface device located in said environment in a manner differing from previously assigned instructions.

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

3. The method of claim 1, wherein said functioning of at least one of said actor, said environment, and said system is selected from the group comprising:

an activity, a preference, a status description, and performance.

4. The method of claim 1, wherein said learned model of behavior provides at least one property selected from the group consisting:

time of day, day of week, day of month, day of year, month, year, duration, frequency, likelihood of activity occurring, likelihood of activity not occurring, likelihood of commencing activity, likelihood of abandoning activity, possible outcomes, and effectiveness of outcomes.

5. The method of claim 1, wherein said learned model of behavior provides at least one property selected from the group comprising:

frequency of correlation between functionings, dependence of the functionings, duration between start time 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.

6. The method of claim 1, further comprising:

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

7. The method of claim 1, further comprising:

providing said learned model of behavior to a second monitoring system that monitors a second human actor in a second environment; and

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

8. 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 second monitoring system that monitors a second human actor in a second environment utilizing said meta-model of behavior as a baseline.

9. The method of claim 1, further comprising:

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

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

10. The method of claim 9, wherein said feedback information relates to a preference of said actor.

11. The method of claim 9, wherein said feedback information relates to performance of said system.

12. The method of claim 1, further comprising:

posing questions; and

recording responses to said questions;

wherein said learned model of behavior is based upon said responses.

13. The method of claim 12, wherein said questions relate to a status of at least one of said actor and said environment.

14. The method of claim 12, wherein said questions relate to an activity of at least one of said actor and said environment.

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

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

comparing said current activity with said expected activity.

16. The method of claim 1, further comprising:

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

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

condensing multiple, sequential firings of said at least one sensor into a single data item.

18. The method of claim 1, wherein said at least one sensor comprises a plurality of sensors, the method further comprising:

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

19. The method of claim 1, further comprising:

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

20. The method of claim 19, wherein said system includes an intent recognition module adapted to determine an intended activity of said actor based upon a series of sensed actions, and further wherein assessing a said current situation includes:

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

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

21. The method of claim 19, wherein said learned model of behavior is indicative of steps normally performed by said actor in completing a particular task, and further wherein assessing a said current situation includes:

determining whether said actor is engaged in said particular task;

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

determining whether said actor has abandoned said particular task based upon said comparison.

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

23. The method of claim 1, wherein said learned model of behavior is indicative of an expected activity, and further wherein operating said system includes generating a response when a current activity varies from said expected activity, and delivering said response to said actor.

24. The method of claim 1, wherein operating said system to generate a response includes:

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

25. The method of claim 24, wherein said 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 said effectiveness of previously generated responses.

26. The method of claim 25, wherein said response includes a message component for providing a message to at least one of said actor and other humans, and further wherein said learned model of behavior is indicative of an effectiveness of previous messages, and further wherein formatting said response includes:

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

27. The method of claim 1, wherein said learned model of behavior is indicative of said actor selecting a particular action of a plurality of action options, and further wherein operating said system to generate said response includes:

generating said response based upon a probability of action selection.

28. The method of claim 1, wherein said learned model of behavior is indicative of a likelihood of said actor successfully completing an activity, and further wherein operating said system to generate response includes:

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

29. The method of claim 1, wherein said learned model of behavior is indicative of an expected path the actor normally follows in said environment, and further wherein operating said system includes:

generating said response based upon said expected path.

30. The method of claim 29, wherein said response includes actuating at least one component of said environment along said expected path.

31. The method of claim 1, wherein automatically generating said learned model of behavior includes:

organizing said probability density function determined therein into sequential patterns.

32. The method of claim 31, wherein a plurality of sensors are provided, and further wherein organizing said probability density function therein includes:

designating discrete partitions for each probability density function for each sensor of said plurality of sensors; and

categorizing said each probability density function for each sensor of said plurality of sensors in accordance with said designated partitions;

wherein said sequential patterns are formulated based upon said categorized probability density functions.

33. The method of claim 32, wherein said discrete partitions are time interval values, and further wherein designating said time interval values includes:

determining at least one time interval value for each said each probability density function for each sensor of said plurality of sensors.

34. The method of claim 31, wherein automatically generating said learned model of behavior further includes:

eliminating redundant sequential patterns from the learned model of behavior.

35. The method of claim 34, wherein eliminating said redundant sequential patterns further includes:

identifying sequential sub-patterns; and

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

36. 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 said at least one sensor to monitor said actor and for controlling operations of said effector based upon the said information, said controller further being adapted to;

accumulate information signaled by said at least one sensor over time;

determine a probability density function for said information signaled by each sensor of said at least one sensor in order to create data indicative of said each sensor activity;

utilize a plurality of filter routines to reduce a quantity of said data indicative of said each sensor activety wherein said plurality of filter routines includes a plurality of the following;

a domain heuristic filter routine, a redundant pattern filter routine, a repetitive sensor firing routine and a cluster technique filter routine;

automatically generate a learned model of behavior relating to one of said actor and said environment based upon said data indicative of said each sensor activity reduced by said plurality of filter routines;

automatically initiate providing of assistance to said actor based upon reference to said learned model of behavior by automatically altering a current mode of operation of said effector in a manner indicated by said learned model of behavior.

37. The system of claim 36, wherein the environment is a daily living environment.

38. The system of claim 36, wherein said learned model of behavior relates to a status of at least one of said actor and said environment.

39. The system of claim 36, wherein said learned model of behavior relates to an activity of at least one of said actor and said environment.

40. The system of claim 36, wherein said learned model of behavior relates to a preference of at least one of said actor and an other person in said environment.

41. The system of claim 36, wherein said learned model of behavior relates to performance of the system.

42. The system of claim 36, wherein said controller is further adapted to:

generate a plurality of learned models of behavior wherein each learned model of behavior relates to a different subject matter.

43. The system of claim 36, wherein said controller is further adapted to:

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

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

44. The system of claim 36, wherein the learned model of behavior is indicative of an expected activity of one of said actor and said environment, and further wherein said controller is adapted to:

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

compare said current activity with said expected activity.

45. The system of claim 36, wherein said controller is further adapted to:

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

46. The system of claim 36, wherein said controller is further adapted to:

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

47. The system of claim 36, wherein said learned model of behavior is indicative of a likelihood of said actor successfully completing a particular activity, and further wherein said controller is adapted to:

recognize whether said actor is engaged in said particular activity; and

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

48. The system of claim 36, wherein said learned model of behavior is indicative of effectiveness of previously generated responses, and further wherein said controller is adapted to:

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

49. The system of claim 36, wherein said controller is further adapted to:

organize said information signaled by said at least one sensor into sequential data patterns.

50. The system of claim 49, wherein said system includes a plurality of sensors, and wherein said controller is further adapted to:

designate discrete partitions for each probability density function for each sensor of said plurality of sensors; and

categorize said each probability density function for each sensor of said plurality of sensors in accordance with the designated partitions;

wherein said sequential patterns are formulated based upon said categorized probability density functions.

51. The system of claim 50, wherein said discrete partitions are time interval values and wherein said controller is further adapted to:

determine at least one time interval value for said each probability density function for each sensor.

52. The system of claim 49, wherein said controller is further adapted to:

eliminate redundant sequential patterns from said learned model of behavior.

53. The system of claim 52, wherein said controller is further adapted to:

identify sequential sub-patterns; and

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

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

receiving a plurality of data items for at least two weeks from a plurality of information sources wherein each information source of said plurality of information sources relates to at least one of said actor or an environment of said actor;

assigning a time of day of occurrence to each data items of said plurality of data items from at least one source of said plurality of information sources;

separately aggregating said plurality of data items from said each information source;

determining a probability density function for said plurality of data items from said each information source in order to create data indicative of said each information source;

utilizing a plurality of filter routines to reduce a quantity of said data indicative of said each information source wherein said plurality of filter routines includes a plurality or the following;

a domain heuristic filter routine, a redundant pattern filter routine, a repetitive sensor firing routine and a cluster technique filter routine;

determining discrete partitions for said each information source of said plurality of information sources based upon an evaluation of said data items and said time of day of occurrence;

assigning a corresponding partition value to said each data item;

automatically generating a learned model of behavior based upon said corresponding partition value; and

utilizing said learned model of behavior thereby created to generate responses by actuators in said environment.

55. The method of claim 54, further comprising:

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

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

providing a plurality of sensors for monitoring a human actor in a daily living environment including a sensor adapted to pose, and record responses to, questions presented to the actor;

accumulating information signaled by said plurality of sensors over time indicative of said human actor functioning in said daily living environment; and

creating a plurality of probability density functions by determining a probability density function for each sensor of said plurality of sensors in order to create data indicative of said each sensor activity;

utilizing a plurality of filter routines to reduce a quantity of said data indicative of said each sensor activity wherein said plurality of filter routines includes a plurality of the following;

a domain heuristic filter routine, a redundant pattern filter routine, a repetitive sensor firing routine and a cluster technique filter routine;

generating a learned model of behavior based upon said data indicative of said each sensor activity reduced by said plurality of filter routines; and

utilizing said learned model of behavior to generate responses by actuators in said environment.