An Adaptive Neuro-Fuzzy Approach for Activity Recognition in Situation-aware Wearable Systems

Wearable computing systems are capable of unobtrusively and continuously monitoring humans by collecting reliable data. Situation-aware human activity recognition, based on wearable systems, is the task of identifying the activities performed by one or more users with respect to the situation in which such activity happens. In this paper, starting from a reference architecture of a situation-aware wearable computing system, an approach for situation-aware human activity recognition is proposed. The approach uses an adaptive neuro-fuzzy inference system to identify the activities by processing wearable sensor data. Such activities are used as the input of a fuzzy inference system, together with other contextual information, to identify the situations involving the users. The approach is evaluated using the MHEALTH dataset, comparing its performance with an artificial neural network technique, showing promising results.