Perception of Human Periodic Motion in Moving Light Displays - a Motion-based Frequency Domain Approach

We present a motion-based frequency-domain technique for modelling and recognising human periodic movements in moving light displays (MLDs). Periodic motions are modelled by motion templates composed of a set of feature power vectors. Feature power vectors are extracted from unidentified trajectories of feature points using motion power spectral analysis. Motion recognition therefore is carried out in the frequency domain by finding a best match among pre-stored templates. Recognition is demonstrated through examples of human periodic motion with data obtained from a marker-based motion capture system. Experimental results prove that motion characteristics exist not only in the spatio-temporal domain, but also in frequency domain. This method contrasts with common spatio-temporal approaches and avoids a time consuming recovery of underlying kinematic structures in visual analysis.