Fall detection using sensor fusion

• US 8,614,630 B2
• Filed: 11/12/2012
• Issued: 12/24/2013
• Est. Priority Date: 11/14/2011
• Status: Active Grant

• Current Assignee: Vital Connect, Inc.
• Original Assignee: Vital Connect, Inc.
• Inventors: Ferdosi, Nima, Narasimhan, Ravi, Chan, Alexander
• Primary Examiner(s): Mehmood, Jennifer
• Assistant Examiner(s): GIRMA, FEKADESELASS
• Application Number:
  US13/674,826
• Publication Number:
  US 20130120147A1
• Time in Patent Office: 407 Days
• Field of Search: 340/573.1
• US Class Current: 340/573.1
• CPC Class Codes: A61B 5/1117 Fall detection
• CPC Class Codes: A61B 5/725 using specific filters therefor, e.g. Kalman or adaptive filters A61B5/316 take precedence
• CPC Class Codes: G01P 15/00 Measuring acceleration; Measuring deceleration; Measuring shock, i.e. sudden change of acceleration
• CPC Class Codes: G01P 15/0891 with indication of predetermined acceleration values G01P15/135 takes precedence
• CPC Class Codes: G08B 21/02 Alarms for ensuring the safety of persons
• CPC Class Codes: G08B 21/043 detecting an emergency event, e.g. a fall
• CPC Class Codes: G08B 21/0446 worn on the body to detect changes of posture, e.g. a fall, inclination, acceleration, gait

Abstract

A method and system for fall detection using sensor fusion are disclosed. In a first aspect, the method comprises in response to any of first and second acceleration magnitude thresholds being satisfied, determining whether a height difference before and after impact of a fall satisfies a threshold and whether an angle threshold between an acceleration vector and a calibration vector is satisfied. In a second aspect, the system comprises a processing system and an application coupled to the processing system, wherein the application carries out the steps of the method.

Claims

1. A method for fall detection, the method comprising:

in response to any of first and second acceleration magnitude thresholds being satisfied, determining whether a height difference before and after impact of a fall satisfies a threshold and whether an angle threshold between an acceleration vector and a calibration vector is satisfied; and

determining whether a maximum value of an activity metric is less than another threshold, wherein if the maximum value of the activity metric is less than the another threshold and the height difference before and after impact satisfies the threshold, the fall detection is confirmed.

2. The method of claim 1, wherein any of first and second acceleration magnitude thresholds being satisfied further comprises:

obtaining an acceleration sample from a user;

comparing the acceleration sample to a first acceleration threshold;

wherein if the acceleration sample is less than the first acceleration threshold, the first magnitude threshold is satisfied, else comparing the acceleration sample to a second acceleration threshold; and

wherein if the acceleration sample is greater than the second acceleration threshold, the second magnitude threshold is satisfied.

3. The method of claim 2, wherein comparing the acceleration sample to the first acceleration threshold further comprises:

applying two filters to the acceleration sample to output an acceleration vector;

calculating L^p-norm of the acceleration vector to output an acceleration scalar; and

comparing the acceleration scalar to the first acceleration threshold.

4. The method of claim 2, wherein comparing the acceleration sample to the second acceleration threshold further comprises:

applying two filters to the acceleration sample to output an acceleration vector;

calculating L^p-norm of the acceleration vector to output an acceleration scalar; and

comparing the acceleration scalar to the second acceleration threshold.

5. The method of claim 3, wherein L^p-norm is any of L¹-norm, L²-norm, L^∞

-norm and the two filters are any of single-pole infinite impulse response (IIR) filters, multiple-pole IIR filters, finite impulse response (FIR) filters and median filters.

6. The method of claim 4, wherein L^p-norm is any of L¹-norm, L²-norm, L^∞

-norm and the two filters are any of single-pole infinite impulse response (IIR) filters, multiple-pole IIR filters, finite impulse response (FIR) filters and median filters.

7. The method of claim 1, wherein determining whether a maximum value of an activity metric is less than a threshold further comprises:

after waiting a predetermined time period, applying an IIR band-pass filter to an acceleration vector of a user to produce an activity level vector of the user;

calculating L¹-norm of the activity level vector to output an activity level scalar;

calculating a moving average of the activity level scalar over a time period; and

comparing the moving average to a predetermined activity level, wherein if the moving average is greater than the predetermined activity level, the maximum value of the activity metric is not less than the threshold and notification information of the fall detection is disregarded.

8. The method of claim 1, further comprising:

in response to any of the first and second acceleration magnitude thresholds being satisfied, waiting a predetermined time period and determining any of whether an acceleration vector of a user is at a predetermined angle to a calibration vector and whether the user is at a stooped posture.

9. The method of claim 8, wherein determining whether an acceleration vector of the user is at a predetermined angle to a calibration vector further comprises:

attaching a wireless sensor device to the user;

determining the calibration vector, wherein the calibration vector is an acceleration vector when the user is vertical;

obtaining at least one acceleration sample from the wireless sensor device;

comparing the at least one acceleration sample to the calibration vector; and

wherein if the at least one acceleration sample is nearly orthogonal to the calibration vector, detecting the fall of the user.

10. The method of claim 9, wherein determining the calibration vector further comprises:

attaching a wireless sensor device when the user is vertical; and

measuring an acceleration sample after attachment, wherein the acceleration sample is determined to be the calibration vector.

11. The method of claim 9, wherein determining the calibration vector further comprises:

measuring an acceleration sample after the user is walking, wherein the acceleration sample is determined to be the calibration vector.

12. The method of claim 8, wherein determining whether the user is at a stooped posture further comprises:

calculating a z-axis component of the calibration vector; and

comparing the z-axis component of the calibration vector to a predetermined level.

13. The method of claim 1, further comprising:

relaying notification information of the fall detection of the user to another user or device.

14. The method of claim 1, wherein the height difference is measured by an altimeter, further comprising:

confirming a fall if any of another height difference measured by the altimeter before and after impact satisfies a first threshold for general falls, a height difference satisfies a second threshold for falling down stairs, and a user is lying down for a predetermined time period and an acceleration magnitude is greater than a threshold for falling out of bed.

15. A wireless sensor device for fall detection, the wireless sensor device comprising:

a processing system; and

an application to be executed by the processing system, wherein the application;

in response to any of first and second acceleration magnitude thresholds being satisfied, determines whether a height difference before and after impact of a fall satisfies a threshold and whether an angle threshold between an acceleration vector and a calibration vector is satisfied; and

determining whether a maximum value of an activity metric is less than another threshold, wherein if the maximum value of the activity metric is less than the another threshold and the height difference before and after impact satisfies the threshold, the fall detection is confirmed.

16. The wireless sensor device of claim 15, wherein the application further:

determines whether an acceleration vector of a user is at a predetermined angle to a calibration vector;

determines whether the user is at a stooped posture, and

determines whether a maximum value of an activity metric is less than a threshold.

17. The wireless sensor device of claim 16, wherein the application determines whether any of the first and second acceleration magnitude thresholds are satisfied via an accelerometer coupled to the processing system, and wherein the application determines the height difference via an altimeter coupled to the processing system.

18. The wireless sensor device of claim 16, wherein the application determines whether a maximum value of an activity metric is less than a threshold further comprises that the application:

after waiting a predetermined time period, applies an IIR band-pass filter to an acceleration vector of a user to produce an activity level vector of the user;

calculates L¹-norm of the activity level vector to output an activity level scalar;

calculates a moving average of the activity level scalar over a time period; and

compares the moving average to a predetermined activity level, wherein if the moving average is greater than the predetermined activity level, the maximum value of the activity metric is not less than the threshold and notification information of the fall detection of the user is disregarded.

19. The wireless sensor device of claim 15, wherein the application further:

relays notification information of the fall detection of a user to another user or device.