A software architecture for generic human activity recognition from smartphone sensor data

In these last few years, several recent studies have demonstrated the possibility to perform Human Activity Recognition (HAR) by smartphone sensor data, enabling in this way a new generation of context-aware mobile applications. Smartphone-based HAR systems can exploit the full set of embedded sensors beside the accelerometer in order to increase the accuracy of the detection process. At the same time, the practical deployment of such systems can result highly challenging since it must cope with the limited computational resources and the battery constraints of the mobile devices. In this paper, we address such issues by proposing a novel, generic HAR architecture for Android devices. The system design takes into account the energy constraints by: (i) limiting the number of sensors/features used for the recognition process, while still guaranteeing satisfactory performance in terms of activity detection; (ii) allocating the most CPU intensive tasks, like the training and data processing phases, on an external server. At the same time, the system operations automatize the full learning process, simply notifying the user when a new classification model has been installed on the client. We validate the proposed HAR systems on two use-cases, i.e. transportation mode detection and walking mode detection, and we describe an application for indoor floor detection. Measurements show that the overall accuracy of the activity recognition process can be up to 90% for both the use-cases.