Ambulation : a tool for monitoring mobility over time using mobile phones Computational Science and...
-
Upload
edmund-cox -
Category
Documents
-
view
213 -
download
0
Transcript of Ambulation : a tool for monitoring mobility over time using mobile phones Computational Science and...
Ambulation : a tool for monitoring mobility over time using mobile phones
Computational Science and Engineering, 2009. CSE '09. International Conference
Jason Ryder et al.
Introduction
• Health monitoring systems– Cost-prohibitive and complicated– Less expensive and simpler alternative– By using mobile phone, infer user’s activity– Logged data is uploaded to server– User is able to login to web server and view data
• Reduce energy usage– Intelligent use of GPS
2
Related Works : Monitoring the health of patients at home
• Attentive Care – Video observation that help care giver to care the care
receiver– Does not track the user’s motion automatically
• Quiet Care– Deploy motion sensors around the care receiver’s home– Analyze changes in the amount of daily walking– Requires specialized sensors and base station– Does not work if the patient moves away from the sensor
• WellAWARE– Similar to Quiet care, but use several kinds of sensors– Capture more type of data– Requires specialized sensors and base station
3
System architecture
• System architecture
4
Classification
• Mobile Sensing Client – Record acceleration and speed information– Run classifier
• Indoor Classifier– Collect accelerometer data– Use decision tree– It can detect stationary, walking, running state
• Outdoor Classifier– Collect accelerometer and GPS data– It can detect still, walking, running, biking, or driving
5
Mobile Sensing Client
GPS Power Optimization
6
Mobile Sensing Client
GPS Power Optimization
• Test data : – 100 trips ,116hours ,25days, 80 hours stationary, 36 hours mov-
ing– Recorded in urban and suburban– Involved walking and freeway driving
• Energy saving depend on how mobile the user is
• Battery lifetime– 6.3 hours 9 hours (in typical scenario), 43% gain
7
Mobile Sensing Client
Data Upload Power Optimization
• Naïve approach– User activate and deactivate upload manually– User may forget to en/dis-able
• Potentially long data upload delays• Quickly depleted battery
• Approach in this study– Upload data upon application start up– Upload only when plugged into an external power source
8
Mobile Sensing Client
Ambulation summary visualization
9
Mobile Sensing Client : Processing module and Visualization
Daily trace calender
10
Mobile Sensing Client : Processing module and Visualization
Conclusion
• Ambulation : Health monitoring system– Indoor and outdoor classifier– Adaptive GPS strategy save power– Upload collected mobility information to server– Visualization
• Future work : Automated anomaly detection– Use mature statistical algorithms– Eigen behavior create similar score for different
datasets
11