Localization of Humans, Objects, and Robots Interacting on Load-Sensing Floors

摘要: Localization, tracking, and recognition of objects humans are basic tasks that high value in the applications ambient intelligence. Sensing floors were introduced to address these a non-intrusive way. To recognize moving on floor, they usually first localized, then set gait features extracted (stride length, cadence, pressure profile over footstep). However, generally fails when several people stand or walk together, preventing successful tracking. This paper presents detection, technique which uses objects’ weight. It continues working even tracking individual persons becomes impossible. Inspired by computer vision, this processes floor pressure-image segmenting blobs containing objects, them, recognizing their contents through mix inference combinatorial search. The result lists probabilities assignments known observed blobs. concept was successfully evaluated daily life activity scenarii, involving multi-object low-resolution sensors, crossing user trajectories, weight ambiguity. can be used provide probabilistic input for multi-modal object systems.