UJA Human Activity Recognition Multi-Occupancy Dataset.

Activity Recognition Systems ARS are proposed to improve the quality of human life. An ARS uses predictive models to identify the activities that individuals are performing in different environments. Under data-driven approaches, these models are trained and tested in experimental environments from datasets that contain data collected from heterogeneous information sources. When several people interact (multi-occupation) in the environment from which data are collected, identifying the activities performed by each individual in a time window is not a trivial task. In addition, there is a lack of datasets generated from different data sources, which allow systems to be evaluated both from an individual and collective perspective. This paper presents the SaMO – UJA dataset, which contains Single and Multi-Occupancy activities collected in the UJAmI (University of Jaén Ambient Intelligence, Spain) Smart Lab. The main contribution of this work is the presentation of a dataset that includes a new generation of sensors as a source of information (acceleration of the inhabitant, intelligent floor for location, proximity and binary-sensors) to provide an excellent tool for addressing multioccupancy in smart environments.