radio sensing: a monitoring technology for resilient places...raspberry pi phillip lundrigan, kyeong...
TRANSCRIPT
Radio Sensing: A Monitoring Technology for Resilient Places
Dr. Neal Patwari 26 April 2017
Research area: Radio Sensing
The wireless is the sensor
“The number of bars” “Signal strength”
How it works: Location from radio sensing
1. Plug in sensors and place them on a map
2. They form a mesh network covering area
3. Measure changes on each link line
4. Motion is detected and shown on the map
Accurate to about 1 meter Can be hidden See through objects Not video or audio
surveillance Real-time access Easy deployment Low cost ($500 / 1500 ft2)
Product: Home motion location sensor
http://xandem.com
Track wakeup time Translate tracking to activity class (meal prep, bathroom) Estimate speed of walking (look for changes over months) Collect data simultaneously from other sensors Detect and track multiple people Distinguish pet via radio tag on collar
In the lab: Extensions for health monitoring
O. Kaltiokallio, R. Jäntti, and N. Patwari, ARTI: An adaptive radio tomographic imaging system, IEEE Trans. Vehicular Technology, Jan 2017. O. Kaltiokallio, M. Bocca, and N. Patwari, Follow @grandma: long-term device-free localization for residential monitoring, IEEE SenseApp, Oct. 2012.
Two sensors Body inhalation & pulse changes multipath waves → Signal strength (sum of waves) changes
Radio waves
Setup: In-home non-contact RR / HR monitor
N. Patwari, L. Brewer, Q. Tate, O. Kaltiokallio, and M. Bocca, Breathfinding: A Wireless Network that Monitors and Locates Breathing in a Home, IEEE J. Selected Topics in Signal Processing, Nov. 2013
Small variations from breathing are monitored
Anh Luong, Alemayehu Solomon Abrar, Thomas Schmid, and Neal Patwari, RSSI step size: 1 dB is not enough!, in Proc. 3rd ACM Workshop on Hot Topics in Wireless, 3 Oct. 2016.
New generation hardware provides increased accuracy
1.9 in Sig
nal
Stre
ngth
(dB
) R
IP B
elt S
igna
l (A
DC
Val
ue)
Time (seconds)
Peaks, valleys of signal strength line up with those measured with a respiratory impedance plethysmo-graphy belt
Breathing change in RSS is small (vs. motion) Average errors of about 1 breath/min (vs. RIP)
Smaller changes from pulse can be monitored
By filtering noise and slow drift, we can see breathing clearly.
Filtering out breathing we see a signal at harmonics of the pulse rate, 10x smaller than breathing
Al. Abrar, et al. Save Our Spectrum: Device-free Human Sensing Using Single Carrier Radio, (submitted) ACM SenSys 2017
Sig
nal
Stre
ngth
(dB
)
Time (seconds) Time (seconds)
Beats Per Second Breaths Per Second S
igna
l S
treng
th (d
B)
Pow
er S
pect
ral
Den
sity
Pow
er S
pect
ral
Den
sity
Pulse estimation accuracy: Experimental Ground truth: Pulse ox Average Error:
1.6 bpm lying 3.7 bpm sitting
Pul
se R
ate
(bre
aths
per
min
ute)
Time (seconds)
User 1 User 2
Al. Abrar, et al. Save Our Spectrum: Device-free Human Sensing Using Single Carrier Radio, (submitted) ACM SenSys 2017
Conclusion
Goals: Find research collaborators / partners Test next-gen system Extend features for home health Enable research Enable consumer systems
We can readily monitor a resident’s: location, breathing rate, pulse rate, without their active participation
Extra Slides
PRISMS In-Home Sensing System Architecture Visualization
Tool
Sensors / Actuators
Gateway
Server
Home Assistant
Local DB e.g.
Influx
BLE WiFi Zwave
Cellular / Home Broadband
Gateway
Raspberry Pi
Phillip Lundrigan, Kyeong T. Min, Jimmy Moore, Sneha Kumar Kasera, Kathy Sward, and Neal Patwari “EpiFi: An In-Home Sensor Network Architecture for Epidemiological Studies”, (submitted) SenSys 2017.
Deployment A: Gateway Internal Data
Location tracking using
smartphones
Hue light bulbs
Z-Wave multi-sensor
Wifi Switch and motion detector, Z-Wave door sensor
WiFi temperature and humidity sensor
Z-Wave energy meter
Utah modified Dylos sensor