A Non-GPS Low-Power Context-Aware System using Modular Bayesian Networks

The proliferation of smartphones has lead to the development of a large variety of applications and the inverstigation on the use of various sensors through contextawareness, in order to provide better services. However, smartphone battery capacity is extremely limited, so that the applications cannot be effectively used. In this paper, we propose a low-power context-aware system using modular Bayesian networks. Bayesian networks are known to respond flexibly to uncertain situations. However, probabilistic models, such as Bayesian networks, have high time complexity, resulting in high power consumption. To reduce the time complexity, we modularize the network based on the Markov boundary, and eliminate the use of GPS because it consumes a lot of power. We compare the accuracy of the system using a combination of sensors and confirm the decrease in the time complexity. Experiments with the real data collected show that the proposed Bayesian networks yield an accuracy of 92.47%. Keywords-Low-power system; context-awareness; modular Bayesian network; Markov boundary.