Using Environmental Sensors to Estimate Users Activity
Using Environmental Sensors to Estimate Users Activity Mike Wang1, Dorothy W. Curtis, M.Sc.2 1MIT EECS –
Undergraduate Research and Innovation Scholar 2Computer Science and Artificial Intelligence Laboratory, MIT, Cambridge, MA,USA
Project Objectives
Wheelchair User Challenges Issues: pressure ulcers, irritation from dermal sensors, environmental exposure, fatigue Solution: wheelchair with non-invasive sensor system to collect vital signs, monitor environmental conditions and user motion Temperature and humidity sensors
I: Analyze data collected by the wheelchair to determine whether a person is inside or outside at a given time
II: Use processed data to create a historical trend summary of activity level
Backrest pressure sensors (BCG) Seat pressure sensors (BCG)
III: Find ways to improve the system
Accelerometer (Arias et al. AMIA Annual Symposium, 2013)
Algorithm Overview Estimate whether wheelchair is indoors or outdoors 1. Wheelchair starts indoors 2. Significant temperature change signals move outdoors 3. Next significant temperature change signals move indoors 4. But, indoor temperature changes over time: thin red line 5. Also, it might be cooler outdoors than indoors: green line
Raw Data
Filtering
Windowing
Inference of inside values
historical data for outside values
Interpolation
Results and Graphs 1
3
2
1 and 2) Results of running the Algorithm on Temperature Data
Conclusions
3) Summary for the 4 wheelchairs over 3 days. Tells how many hours a person has been outside on a given day. NA’s tell us there is not enough data to make an inference.
Future Directions Test trend summary with caregivers and users
I: Determine whether a person is outside at a given time Conclusion: Using historical weather data performs decently well; however, weather data does not seem to reflect outside temperatures very well. II: Create historical trend summary of activity level Conclusion: Aggregated results of algorithm into a table that shows trends in activity levels III. Find ways to improve the system Conclusion: We should also deploy stationary indoor and outdoor temperature sensors
Try indoor and outdoor temperature sensors Try to use accelerometer data Collect truth values for supervised learning techniques Wheelchair prototype testing
Acknowledgements Thanks to the SuperUROP Program for supporting our work through the MIT EECS - Undergraduate Research and Innovation Scholar. Thanks to The Boston Home for supporting this project and to Diego Arias, Esteban Pino, Pablo Aqueveque, and Rui Song for initial work on this project.
Undergraduate Research and Innovation Scholar 2Computer Science and Artificial Intelligence Laboratory, MIT, Cambridge, MA,USA
Project Objectives
Wheelchair User Challenges Issues: pressure ulcers, irritation from dermal sensors, environmental exposure, fatigue Solution: wheelchair with non-invasive sensor system to collect vital signs, monitor environmental conditions and user motion Temperature and humidity sensors
I: Analyze data collected by the wheelchair to determine whether a person is inside or outside at a given time
II: Use processed data to create a historical trend summary of activity level
Backrest pressure sensors (BCG) Seat pressure sensors (BCG)
III: Find ways to improve the system
Accelerometer (Arias et al. AMIA Annual Symposium, 2013)
Algorithm Overview Estimate whether wheelchair is indoors or outdoors 1. Wheelchair starts indoors 2. Significant temperature change signals move outdoors 3. Next significant temperature change signals move indoors 4. But, indoor temperature changes over time: thin red line 5. Also, it might be cooler outdoors than indoors: green line
Raw Data
Filtering
Windowing
Inference of inside values
historical data for outside values
Interpolation
Results and Graphs 1
3
2
1 and 2) Results of running the Algorithm on Temperature Data
Conclusions
3) Summary for the 4 wheelchairs over 3 days. Tells how many hours a person has been outside on a given day. NA’s tell us there is not enough data to make an inference.
Future Directions Test trend summary with caregivers and users
I: Determine whether a person is outside at a given time Conclusion: Using historical weather data performs decently well; however, weather data does not seem to reflect outside temperatures very well. II: Create historical trend summary of activity level Conclusion: Aggregated results of algorithm into a table that shows trends in activity levels III. Find ways to improve the system Conclusion: We should also deploy stationary indoor and outdoor temperature sensors
Try indoor and outdoor temperature sensors Try to use accelerometer data Collect truth values for supervised learning techniques Wheelchair prototype testing
Acknowledgements Thanks to the SuperUROP Program for supporting our work through the MIT EECS - Undergraduate Research and Innovation Scholar. Thanks to The Boston Home for supporting this project and to Diego Arias, Esteban Pino, Pablo Aqueveque, and Rui Song for initial work on this project.
