Human Activity Recognition

Methods of human activity recognition (HAR) have been developed for the purpose of automatically classifying recordings of human movements into a set of activities. Capturing, evaluating and analysing sequential data to accurately recognise human activities is critical for many applications in pervasive computing, for example in applications such as ambient-assisted living, smart-homes, activities of daily living, health support, and industry 4.0. A HAR system seeks to classify the activities of a subject based on the recording of physical quantities related to the subject's movement. The HAR system consists of different stages: data recording, data representation, preprocessing, and classification, either by statistical feature recognition or by neural networks. HAR processes signals from videos, marker-based motion capturing system, or inertial measurement units (IMUs). The latter ones are very important as they make HAR a potential tool beyond constrained or laboratory settings. IMUs are not affected by occlusion, and they do not portray human identities. IMUs are low-power devices that are cheap, highly reliable, non-invasive and easy to use. They should assist anywhere and anytime by observing activities egocentrically. They provide a view into the movement of the subject wearing the IMUs. All of this makes HAR a method of special interest for optimisation in those industries where manual work remains dominant such as logistics.