Zhixian Zan' talk on DB Seminar

 

Title: Energy-Efficient Continuous Activity Recognition on Mobile Phones

Abstract
Power consumption on mobile phones is a painful obstacle towards adoption of continuous sensing driven appli- cations, e.g., continuously inferring individual’s locomotive activities (such as ‘sit’, ‘stand’ or ‘walk’) using the embedded accelerometer sensor. To reduce the energy overhead of such continuous activity sensing, we first investigate how the choice of accelerometer sampling frequency & classification features affects, separately for each activity, the “energy overhead” vs. “classification accuracy” tradeoff. We find that such tradeoff is activity specific. Based on this finding, we introduce an activity-sensitive strategy (dubbed “A3R” – Adaptive Accelerometer-based Activity Recognition) for continuous activity recognition, where the choice of both the accelerometer sampling frequency and the classification features are adapted in real-time, as an individual performs daily lifestyle-based activities. We evaluate the performance of A3R using longitudinal, multi-day observations of continuous activity traces. We also implement A3R for the Android platform and carry out evaluation of energy savings. We show that our strategy can achieve an energy savings of 50% under ideal conditions. For users running the A3R application on their Android phones, we achieve an overall energy savings of 20-25%.

 

Bio

Zhixian Yan is a postdoctoral researcher at the Distributed Information Systems Laboratory (LSIR) at EPFL. He received his Ph.D. in computer sciences at EPFL in August 2011. His research interests include mobile/sensor data management, stream data processing & mining, and various topics towards mobile sensing, such as community/participatory/collaborative sensing, energy-efficient activity learning, and environmental sensing.