Non-parametric modeling apparatus and method for classification, especially of activity state

• US 7,818,131 B2
• Filed: 06/19/2006
• Issued: 10/19/2010
• Est. Priority Date: 06/17/2005
• Status: Expired due to Fees

• Current Assignee: Prolaio, Inc.
• Original Assignee: Venture Gain LLC
• Inventors: Mott, Jack E.
• Primary Examiner(s): Hindenburg; Max
• Assistant Examiner(s): DANEGA, RENEE A
• Application Number:
  US11/455,495
• Publication Number:
  US 20070010720A1
• Time in Patent Office: 1,583 Days
• Field of Search: 600/300, 382/115
• US Class Current: 702/19
• CPC Class Codes: G16H 50/70 for mining of medical data, e.g. analysing previous cases of other patients
• CPC Class Codes: G16Z 99/00 Subject matter not provided for in other main groups of this subclass

Abstract

The activity state classification method of the present invention employs a kernel-based modeling technique, and more specifically a set of similarity-based models, which have been created using example data, to process an input observation or set of input observations, each comprising a set of sensor readings or “features” derived there from or other data, to predict the activity state of a person from whom the sensor data was obtained. A model is created for each class of activity. The input data is processed by each model and the resulting predictions are combined to yield a final prediction of which state of activity is represented by the input data.

Claims

1. A method for determining the class of an activity selected from a plurality of activities of interest, of a person based on sensor data, comprising the steps of:

providing a plurality of kernel-based models, each model corresponding to one of said activities of interest,

obtaining a set of readings of said sensor data;

generating estimates of said set of sensor data readings from at least some of said kernel-based models based on an input of said set of sensor data readings;

comparing for similarity each said estimate of said set of sensor data readings to said input set of sensor data readings, and

determining the class of activity corresponding to said input set of sensor data readings to be the class of the model that generated the estimate most similar to the input set.

2. A method according to claim 1, wherein at least one of said kernel-based models is a similarity-based model.

3. A method according to claim 2, wherein said similarity-based model employs an elemental similarity operator.

4. A method according to claim 1, wherein the step of comparing comprises calculating a global similarity value using a similarity kernel function on the estimate of said set of sensor data readings and said set of sensor data readings.

5. A method according to claim 2, wherein said step of comparing further comprises normalizing said global similarity values according to an expectation of global similarities typically generated by each said kernel-based model.

6. A method according to claim 1, further comprising the step of determining a class of activity of a time series window of sets of readings of said sensor data based on the determinations of the classes of activities for each of the input set of sensor data readings.

7. A method according to claim 1, wherein said sensor data includes a measure of heart rate.

8. A method according to claim 1, wherein said sensor data includes a measure of temperature of the person.

9. A method according to claim 1, wherein said sensor data includes a measure of respiration.

10. A method according to claim 1, wherein said sensor data is transmitted from a device that collects sensor data from the person, to a remote computer having in memory the provided plurality of kernel-based models, and where estimate generation, estimate comparison and activity class determination are performed on said remote computer.

11. A method according to claim 10, wherein said remote computer presents the determined class of activity via a web browser.

12. A method according to claim 10, wherein said remote computer uses said determination of activity class to further determine whether to use the sensor data for a medical assessment of the person.

13. A method according to claim 1, wherein said sensor data is obtained by a device, said device having memory for storing the provided plurality of kernel-based models, and a processor for executing program code that performs estimate generation, estimate comparison and activity class determination.

14. A method according to claim 13 wherein said device uses said determination of activity class to further determine whether to use the sensor data for a medical assessment of the person.

15. A method according to claim 13 wherein said device uses said determination of activity class to determine whether or not to initiate a notification to the person.