APPARATUS AND METHOD FOR ANALYSING EVENTS FROM SENSOR DATA BY OPTIMISATION

• US 20130254141A1
• Filed: 06/16/2011
• Published: 09/26/2013
• Est. Priority Date: 06/16/2010
• Status: Active Grant

• Current Assignee: Fred Bergman Healthcare Pty Ltd.
• Original Assignee: Fred Bergman Healthcare Pty Ltd.
• Inventors: Barda, David Albert, Mashinchi, Mohammad Hadi
• Granted Patent: US 9,224,102 B2
• Time in Patent Office: Days
• US Class Current: 706/12
• CPC Class Codes: A61F 13/42 with wetness indicator or alarm
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: G06N 5/04 Inference or reasoning models
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks

Abstract

The present invention relates to sensor signal analysis. It relates particularly, but not exclusively, to methods, systems and devices for monitoring and processing the sensor signals to determine automatically characteristics of events represented by the sensor signals. The present invention is particularly, but not exclusively, related to methods, systems and devices for monitoring moisture in absorbent articles such as diapers, incontinence garments, dressings and pads resulting from wetness events caused by, for example, urinary and/or faecal incontinence. In an embodiment, the invention includes a method for processing sensor signals representing an event in an absorbent article. The method comprises: receiving sensor signals from a sensor representing one or more events in an absorbent article; and processing the sensor signals to determine a characteristic of at least one event in the absorbent article. One such characteristic can include the volume of a voiding event such as a urinary incontinence event. In another embodiment, the method includes carrying out a learning phase including the steps of: receiving sensor signals representing one or more events in each of one or more absorbent articles; receiving observation data indicative of a cumulative characteristic of the one or more events in each absorbent article; and identifying an optimal mathematical model describing a relationship between the sensor signals and the observation data. Such events can include urinary incontinence events occurring in absorbent articles such as diapers. Observation data can be measured cumulative volume of a cycle of voiding events occurring in a diaper.

Claims

1. A method for processing sensor signals representing an event in an absorbent article, the method comprising:

receiving sensor signals from a sensor representing one or more events in an absorbent article;

processing the sensor signals to determine a characteristic of at least one event in the absorbent article.

2. The method of claim 1, wherein processing the sensor signals includes identifying automatically the sensor signals representative of each individual event and, from the sensor signals representative of each individual event:

(i) generating a representative vector for that event;

(ii) allocating weightings to the representative vector to generate a weighted representative vector for that event; and

(iii) allocating an event characteristic to the weighted representative vector for that event.

3. The method of claim 2, including carrying out a learning phase including the steps of:

receiving sensor signals representing one or more events in each of one or more absorbent articles;

receiving observation data indicative of a cumulative characteristic of the one or more events in each absorbent article;

identifying an optimal mathematical model describing a relationship between the sensor signals and the observation data.

4. The method of claim 3, wherein the mathematical model includes a system of linear equations describing the relationship between the sensor signals and the observation data.

5. The method of claim 3, further including:

(i) generating a representative vector for each individual event, the representative vector being comprised of one or more elements;

(ii) allocating weightings to the elements of the representative vector to generate a weighted representative vector;

(iii) allocating each of the weighted representative vectors to one or more of a plurality of clusters according to their relative similarity; and

(iv) allocating a characteristic to each of the clusters based on the optimal mathematical model obtained during the learning phase.

6. The method of claim 5, further including normalizing the elements of the representative vector with respect to a reference range and allocating the weightings to the normalized elements of the representative vector to generate a weighted normalized representative vector.

7. The method of claim 5, further including categorizing the representative vectors of events according to order of occurrence in a sequence of events in the absorbent article and allocating different weightings to elements of the representative vectors according to their category.

8. The method of claim 5, wherein the elements of the representative vector include one or more values derived from the sensor signals.

9. The method of claim 8, wherein the values derived from the sensor signals includes any one or more of the group including:

duration of the event,

time to reach a maximum sensor signal value for the event,

average value of the sensor signal during an increase in the signal to a maximum for the event,

the maximum signal value for the event,

the time taken to reach a minimum signal value for the event,

average value of the sensor signal during a decrease in the signal to a minimum for the event,

the minimum signal value for the event,

the order of the event in a sequence of events,

the similarity of a single sensor signal to one or a combination of other sensor signals occurring at the same time.

10. The method of claim 5, wherein the characteristics allocated to each of the clusters includes event information.

11. The method of claim 10, wherein the characteristics allocated to each of the clusters includes a reference representative vector and a void volume.

14. The method of claim 5, further including determining degree of belief of belonging information including, for each of the weighted representative vectors, determining a degree of belief of belonging to one or more of the plurality of clusters.

15. The method of claim 14, further including training a fuzzy neural network with the weighted representative vectors and the degree of belief of belonging information determined for each of the weighted representative vectors, and allocating information to each of the clusters including a void volume and a series of values corresponding to the degrees of belief of belonging information determined for each of the representative vectors.

17. The method of claim 15, further including using the trained fuzzy neural network during an assessment phase to determine degree of belief of belonging information for a weighted representative vector representing an event in an absorbent article including determining a degree of belief of belonging to one or more of the plurality of clusters.

20. The method of claim 3, further including verifying the correctness of the optimal mathematical model by receiving one or more sensor signals from a sensor representing an event in an absorbent article, processing the sensor signals to determine a characteristic of the event based on the optimal mathematical model obtained during the learning phase and comparing the determined characteristic with observation data.

21. The method of claim 3, wherein identifying an optimal mathematical model includes determining a plurality of objective functions of the mathematical model and determining optimal values for the objective functions in a predetermined hierarchical order.

28. A method of verifying the effectiveness of absorbent article and sensor combinations for providing sensor signals accurately representing events in the absorbent article for use in the method of claim 1, the method including:

receiving sensor signals representing a plurality of events in a plurality of absorbent article and sensor combinations of a first set;

receiving sensor signals representing a plurality of events in a plurality of absorbent article and sensor combinations of a second set;

receiving data describing a known characteristic of each of the events in the absorbent articles of the first and second sets;

identifying for each sets of the absorbent article and sensor combinations an optimal mathematical model describing a relationship between the sensor signals and the observation data,

verifying the optimal mathematical models for the sets absorbent article and sensor combinations by receiving one or more sensor signals from second groups of the first and second sets of absorbent article and sensor combinations representing events in the absorbent articles, processing the sensor signals to determine a characteristic of the events based on the optimal mathematical models, receiving data describing a known characteristic of each of the events in the second groups of the first and second sets of absorbent article sensor combinations and comparing the determined characteristic with the known characteristic of each of the events in the second groups of absorbent articles of the first and second sets.

12. (canceled)

13. (canceled)

16. (canceled)

18. (canceled)

19. (canceled)

22. (canceled)

23. (canceled)

24. (canceled)

25. A method of processing sensor signals representing an event in an environment, the method comprising:

receiving one or more sensor signals representing one or more events in an environment;

processing the sensor signals to determine a characteristic of at least one event in the environment.

26. The method of claim 25, wherein processing the sensor signals includes identifying automatically the sensor signals representative of each individual event and, from the sensor signals representative of each individual event:

(i) generating a representative vector for that event;

(ii) allocating weightings to the representative vector to generate a weighted representative vector for that event; and

(iii) allocating a characteristic to each weighted representative vector for that event.

27. The method of claim 25, including carrying out a learning phase including the steps of:

receiving sensor signals representing one or more events in each of one or more environments;

receiving observation data indicative of a cumulative characteristic of the one or more events in each environment;

identifying an optimal mathematical model describing a relationship between the sensor signals and the observation data.

29. (canceled)

30. (canceled)