Fuzzy framework for activity recognition in a multi-occupant smart environment based on wearable devices and proximity beacons

摘要

A fuzzy framework is proposed for activity recognition in a multi occupancy smart environment based on wearable devices and Bluetooth Low Energy (BLE) beacons. In this context, there is a set of Received Signal Strength Indicators (RSSIs), which is received by wearable devices worn by the inhabitants, from a set of BLE beacons deployed in objects within a smart environment. The RSSIs generated by the BLE beacons are influenced by external factors in radio waves, which are presented in the RSSI as important fluctuations. The main contribution in this work has been to propose the use of fuzzy logic and the fuzzy linguistic approach to model indicators of fluctuation. A new framework is introduced that includes a Mamdani fuzzy inference system to recognize the activity that is being carried out in real time for each inhabitant based on data gleaned from their wearable device. The fuzzy rules in addition to the fuzzy linguistic terms used to model the RSSIs can be customized for each inhabitant according to their habits, customs or routines. A case study of the proposed fuzzy framework in the smart lab of the University of Jaen is presented to demonstrate the effectiveness of the approach.