Human activity mining in multi-occupancy contexts based on nearby interaction under a fuzzy approach

Multioccupation encompasses real-life environments in which people interact in the same common space. Recognizing activities in this context for each inhabitant has been challenging and complex. This work presents a fuzzy knowledge-based system for mining human activities in multi-occupancy contexts based on nearby interaction based on the Ultra-wideband. First, interest zone spatial location is modelled using a straightforward fuzzy logic approach, enabling discriminating short-term event interactions. Second, linguistic protoforms use fuzzy rules to describe long-term events for mining human activities in a multi-occupancy context. A data set with multimodal sensors has been collected and labelled to exhibit the application of the approach. The results show an encouraging performance (0.9 precision) in the discrimination of multiple occupations.