Monitoring Walking Behavior with Mobile Computing

The crux of this dissertation is to understand human walking behavior. Walking is one of the most fundamental human activities. Study of walking behavior is difficult because it demands ubiquitous monitoring of a large subject pool for an extended period of time and consideration of multiple covariates. This research examines three factors that may significantly affect walking activity.

The first factor is self-awareness. To this end, we use a smartphone app (iBurnCalorie) that transforms accelerometer readings to caloric values. The app keeps the user informed about her/his walking output. The original caloric transformation assumes that the users walk on a flat ground, which is not always true. In our research, we propose a novel calibration method that takes into account surface inclination, thus improving the accuracy of the initial mapping.

The second factor that likely affects walking behavior is weather. The most common pollutant is ozone, and walking under unsafe ozone conditions is unhealthy. The current ozone alert technologies are neither highly dynamic nor highly specific in terms of locale and, hence, not very informative for walkers. In this dissertation, we have designed, developed, and evaluated an app called OzoneMap that delivers spatio-temporal ozone information. In a study that we performed, we documented the relevance and usefulness of such ozone information for walkers. Subsequently, we incorporated the OzoneMap app within the iBurnCalorie app.

The third factor that likely affects walking behavior are the role models to which people compare themselves. In fact, the iBurnCalorie app supports the search and selection of role models from its user base; the users can monitor themselves and role models they freely choose. We analyzed entrainment effects among dominant and non-dominant nodes in the app's walker network. Identifying positive and negative reinforcement patterns that occur naturally will inform future interventions where struggling walkers, based on their characteristics, will be optimally matched to role models.

The outcome of this work will be an updated iBurnCalorie app that computes caloric expenditure from walking more accurately, maximizes the opportunities for safe walking in polluted metropolitan centers, and virtually pairs walkers with role models that stand the best likelihood of behavioral inducing modification.