Regresar

Noise-sensing using smartphones: Determining the right time to sample

Abstract:

In this paper, we propose to use smartphones as an environmental sensor to measure noise pollution. We focused our study on determining the precise context for the capture of environmental noise through smartphones, and performed an analysis of the impact that the sensing task collection will have on energy consumption. To this purpose, we define different contexts to determine whether adequate environment sampling conditions are met, and we then apply classification algorithms to generate the most accurate decisions trees automatically. An analysis of resource consumption requirements associated with the different trees obtained shows that, despite their high accuracy, the resource consumption levels were prohibitive for this kind of applications. Thus, we propose an alternative decision tree that maintains the accuracy levels of automatically generated trees while significantly reducing the resource consumption introduced by the latter. Experimental results show that our proposed decision tree can reduce the energetic impact of our target application by about 60% when compared to the optimum theoretical tree generated through automatic classification procedures.