Massachusetts Institute of Technology
Enabling Population-Scale Physical Activity Measurement on Common Mobile Phones
Stephen S. Intille
The primary aim of this project is the technical development, deployment, and evaluation of hardware and software technology that will enable population-scale, longitudinal measurement of physical activity and sedentary behavior using common mobile phones. The fast processors and wireless capabilities of common mobile phones make it possible to use these devices with external motion sensors placed on the upper and lower body for accurate measurement of physical activity type, intensity and bout duration. Our project aims to demonstrate that researchers may exploit consumer expenditures on phones that Americans are purchasing anyway to enable large-scale studies where the physical activity of large numbers of participants may be measured and remotely monitored for months or years at an affordable cost.
Using an iterative, participatory design process, we have developed the Wockets motion measurement system (http://web.mit.edu/wockets). This system consists of miniature wireless accelerometers that are optimized to meet the needs of both researchers and subjects. The sensors are:
Multiple sensors transmit raw 3-axis acceleration data to Bluetooth-enabled mobile phones in one of two modes: a continuous mode and a power-efficient “burst” mode. In the burst mode, the system permits participants in an experiment to use their phones normally and wear two sensors for 24 hours and swap them once a day in the morning with a second set of sensors that have been charging overnight.
Participants get into the habit of wearing the comfortable, discrete sensors continuously, under the clothing. Software on the phones can save the raw data for later analysis and/or run pattern recognition algorithms to detect activity type or intensity using real-time analysis of the raw signals from multiple accelerometers and pattern recognition algorithms. Further, the system uses the phone’s data network to upload activity summary data to a secure web server once an hour, and raw data nightly.
A web interface permits remote monitoring of data quality daily by researchers while the sensors are deployed in the field, and participants are automatically messages about their daily compliance. Surveys can also be set up on the phone to gather self-report information to supplement the sensor data.
The Wockets sensor kit consists of:
- Four Wocket sensors,
- Custom moisture-wicking bands, and
- A charger can be manufactured for less than the retail cost of a single Actigraph.
All hardware designs and software are open source and available for others to use. A modified commercial heart rate monitor can also be used with the system. The sensors work with the Android mobile phones.
In small scale laboratory studies, we have validated the sensors’ ability to estimate energy expenditure and detect activity type (with additional experiments analyzing collected data ongoing). With our research team, we have validated that our current system can be used for continuous data collection and remote monitoring of people in the field, and we are now enrolling participants for the final field validation where participants will wear the sensors continuously for up to 8 months. Validity relative to self-report, acceptability, and longitudinal compliance will be measured.
We are working with other investigators interested in using the Wocket system in larger studies. A hardware manufacturer has agreed to inventory the devices. We anticipate data collected in our validation study will provide evidence of the utility and viability of the system for long-term measurement of physical activity and sedentary behavior.