Learning Spatio-temporal Models of Object Behaviour

Many object motions in nature are cyclic and this provides a strong cue for object and activity recognition. Recently, there has been a lot of research effort into detecting and recognising cyclic or near-periodic activities such as a person walking or waving a hand. The main characteristic these approaches share is that they don’t require an object or motion specific model. Periodic and cyclic activities are detected and recognised without knowledge of the underlying object since repetitive motions share common temporal features that can be represented and recognised regardless of the particular object that is moving.