Multi-sensor Based Low Power Sensing Model for Inferencing Location Context

Context-aware applications in mobile devices use various sensors for inferring context to provide high-quality services. However, mobile devices have limited battery life and sensors consume much power. There have been many approaches for energy efficient sensor usage, but specified sensor operations considering context-aware application give more chance to decrease the power consumption of sensor by selecting the most efficient way to infer the context. For specifying sensor operation, we need to define each sensor model enough to fine-grained, but previous studies provide sensor model that is too coarse-grained states. In this paper, we propose fine-grained single sensor model and the method that merges sensor models into multi-sensor based context inference model. We designed framework that manages GPS and Wi-Fi to avoid wasting states. We evaluate our models by applying to the application that infers indoor/outdoor context using GPS and Wi-Fi. Adjusting this model can reduce additional power about 60% decrease for avoided states.