A Low-cost Home Monitoring Architecture for Human Activity Recognition Using Smartwatches

This paper describes the networking architecture of a smartwatch-based monitoring system for biosignal tracking and human activity recognition in home environments. To deploy the application on low-end cost-effective smartwatches and to avoid the high consumption of wider range transmission technologies, the data from the smartwatches are transmitted via Bluetooth Low Energy. In contrast with other similar projects, which require the user to transport an extra gateway node (e.g. a smartphone), our architecture is grounded on a low-cost ‘roaming’ mechanism supported by a scalable network of embedded platforms. A specific central node is in charge of optimizing the selection of the gateway at each moment depending on the signal level with which the data from the watch are received. Thus, smartwatches (or smartbands) become the only wearable of the networking application. The ongoing prototype has been initially employed to implement a basic fall detection application.