Abstract for NIH Grant U01 HL09173

Enabling Population-Scale Physical Activity Measurement on Common Mobile Phones

The following describes the goals of the new 4-year project that the successful applicant would join as a key team member.

Project Abstract:

The goal of this NIH-funded project is the iterative technical development and testing of a hardware and software system that will enable researchers to affordably measure physical activity (PA) type and intensity and its location in very large, free-living populations of adults using ordinary mobile phones. The system we will develop will have four components: (1) one or more miniature wireless accelerometers worn conveniently and comfortably on the body that send accelerometer data to common mobile phones, (2) software that runs on accelerometer-enabled mobile phones that automatically detects PA profile in real-time, (3) software running on the mobile phone that helps participants maintain and use the technology for extended time periods, and (4) server-side web software that allows remote study enrollment, training, data collection, and monitoring of PA in population-scale studies using the sensor and phone technology. The specific aims, in chronological order, are:

To use existing wireless sensors developed at MIT and tested at Stanford (with NSF and NIH funding) to measure the relative information gain about PA intensity and type that can be obtained by wearing one or more 3-axis wireless accelerometers on convenient body locations (e.g., in a watch, attached to a mobile phone, on a shoe, in the pocket, in a purse, etc.). Practical configurations of accelerometers will be compared to determine (a) which configurations can be used by machine learning algorithms to automatically and reliably detect specific PAs such as brisk walking, running, cycling, climbing stairs, sweeping, playing a sport, etc. in real-time, and (b) the type and amount of user-specific training data needed (if any) to obtain good performance.
To determine the technical specifications for a low-cost system of miniature wireless mobile sensors that work with common accelerometer-enabled mobile phones in order to accurately measure PA type and intensity, overcoming the computing and power limitations of the phone and sensor devices by selectively transmitting summary information from the remote sensors and fusing this with continuously-sampled motion data from the phone’s internal sensors, such as accelerometers and GPS.
To iteratively develop and test a power-efficient version of the miniature accelerometers capable of transmitting motion data to phones on the market in the U.S. by 2010 and that address practical end-user concerns about ergonomics, comfort, and social acceptability.
To use participatory design practices with a small panel of phone users to generate, prototype, and implement strategies for obtaining longitudinal compliance with the proposed PA measurement system.
To create software for common mobile phones that uses data from the accelerometers to measure PA profiles in real-time.
To develop software to allow for low-cost, remote maintenance of long-term PA measurement studies using the data communication capabilities of the phone.
To test the system with 50 phone users for 10 months. In the first 5 months, validity of the system’s identification of the frequency, duration, and intensity of specific activities commonly performed by free-living individuals relative to self-report will be measured. In the remaining 5 months, long-term compliance and acceptability will be measured.

By exploiting consumer expenditures on phones that many Americans will purchase, maintain, and carry, the technology may enable PA measurement studies with a high degree of temporal resolution at population scales. An easy process for ordering the hardware will be established, and the software will be free. Both the hardware and software will be open source so that companies, other research teams, or individuals will be able to use and enhance it as needed when this funding expires, including extending the software to save and use data acquired from other devices developed as part of the Exposure Biology Program, a part of the U.S. Genes and Environment Initiative (http://www.gei.nih.gov/exposurebiology/).

Principal Investigator: Stephen Intille, Ph.D.
Institutions: MIT and Stanford Medical School

Last modified: 8/22/07