Towards an Efficient Implementation of Human Activity Recognition for Mobile Devices

Abstract

The availability of diverse and powerful sensors embedded in modern Smartphones/mobile devices has created exciting opportunities for developing context-aware applications. Although there is good capacity for collecting and classifying human activity data with such devices, data pre-processing and model building techniques that achieve this goal are required to operate while meeting hardware resource constraints, particularly for real-time applications. In this paper, we present a comparison study for HAR exploiting feature selection approaches to reduce the computation and training time needed for the discrimination of targeted activities while maintaining significant accuracy. We validated our approach on a publicly available dataset. Results show that Recursive Feature Elimination method combined with Radial Basis Function Support Vector Machine classifier offered the best tradeoff between training time/recognition performance.