The current US population of those 65 years and older is over 40 million today, and the federal Census Bureau predicts that will more than double, to nearly 87 million, by midcentury. There are many individuals who are concerned about and/or are trying to remain in their homes, a class that is being called Aging in Place.
This product idea grew out of discussions with many people about the concern of their parents or relatives growing older and insisting on living alone in the family home – the Aging in Place population. One of the major concerns in this situation is that an individual has a medical episode, a fall or some other physical problem that will not be discovered for some period of time, which would turn the episode from a minor incident into a major problem.
This is where the idea for ‘eCompanion’ was born. With the currently available technology in sensors for movement / position recognition, wireless interconnectivity and mobile devices, a solution is possible where an individual can be monitored and anomalies detected that can trigger contacting a monitoring service without the need of either a 24 visitation service or pushing a button.
With 3axis accelerators monitored in a worn band, a database of the individual's typical movements will be used to provide input for lapses in movement or movements out of normal pattern. A database with adaptive learning capability will be used to input and to determine the typical movement patterns of the monitored individual and convert this into useful information about their daily activities. When the system detects a break in the normal pattern it can automatically trigger an alert sequence.
2. Architecture
Major Components:
• Sensing band worn by MI
• Tethered smartphone
• Cloud based data storage /
learning algorithm
• Call support center
• Resource individual [or Caregiver]
• 911 service as backup
User
database
911 Service
Data
Data
Tethered
Smartphone
Sensor Band
Learning
algorithm
Call
Support
Resource
Individual
3. Sensing Band
Optical heart rate sensor
3-axis accelerometer
Gyrometer
Ambient light sensor
UV sensor
Skin temperature sensor
Capacitive sensor
Galvanic skin response
Microphone
Haptic vibration motor
Bluetooth LE (Low Energy)
Other possible options:
Currently using
Microsoft
Health Band:
4. Tethered
Smartphone
Used to push / pull data from sensor
band to / from cloud based algorithm
system.
Nominal Requirements:
• Platform: Windows / Android
• Processor: Qualcomm Snapdragon
• GPS: A-GPS & GLONASS
• WiFi: Wi-Fi 802.11 b/g/n
• 3G cellular connectivity
5. Fall Detection by
Vector Analysis
MI fall is detected by continually
reading the band’s 3-axis
accelerometer and gyrometer to
provide a six axis look at out of normal
movement patterns.
Since no absolute reference exists,
variations between incremental
readings are compared on timed
intervals for excessive movement that
indicates a fall pattern.
Time Accel X (g) Accel Y (g) Accel Z (g)
Gyro X
(rad/sec)
Gyro Y
(rad/sec)
Gyro Z
(rad/sec)
0.921738808 -0.008774 -0.969223 -0.242905 -0.034438 0 0
0.921739109 -0.012314 -0.971085 -0.236801 -0.034946 0 0
0.921739340 -0.014008 -0.970306 -0.234909 -0.03497 0 0
0.921739711 -0.010468 -0.969437 -0.234192 -0.050642 0 0
0.921739896 -0.011032 -0.969849 -0.252548 -0.048254 0 0
0.921740081 -0.012665 -0.968796 -0.261963 -0.038064 0 0
0.921740313 -0.013107 -0.966553 -0.27182 -0.03409 0 0
0.921740486 -0.013107 -0.966553 -0.27182 -0.031265 0 0
0.921740741 -0.010574 -0.967148 -0.261002 -0.032383 0 0
0.921740926 -0.012222 -0.969315 -0.252792 -0.043219 0 0
0.921741100 -0.009903 -0.969162 -0.251129 -0.024979 0 0
0.921741424 -0.011353 -0.969131 -0.243607 -0.026266 0 0
0.921741667 -0.014862 -0.970001 -0.246445 -0.034849 0 0
0.921741840 -0.009201 -0.967194 -0.256058 -0.032421 0 0
0.921742072 -0.010162 -0.968918 -0.247589 -0.03133 0 0
Readings are compared real time
looking for patterns that fit learning
algorithm developed that model
individual’s projected fall sequence.
6. Automatic
Algorithm
Response
Basic response function based on
analyzing vector movements of band
against machine learned algorithm
adjusted against specific MI movement
patterns
Band event
detected
Band being
worn [HR]
db indicates
escalation
Band dropped
warning
No No
Yes
Robo Alert to
Band
Band Alert
Timeout
Robo Call to
phone
MI responded
Escalate to Call
center
Yes
Event Counter
No response
Local reset
No response
MI OK
7. Call Support
Response
Sequence based on Caregiver’s
instructions on time of day contact
allowances and strength of algorithm’s
response, before direct contact to MI
and hand of to 911
Notification from
Algorithm
Check band
data feed
Time allows
CareGiver
contact
Low priority
Alert band /
call MI
Questionable Yes
Solid
Call CareGiver
CareGiver
Responce
Call MI
MI responded Escalate to 911
No
Timer for
escalation of
response
No response
Responded
No response
MI OKReset / Store
sequence
Timeout